[gentoo-user] nvidia warning comes a tad late

[gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 2110 bytes --]

So, like a good gentoo user I'm emerging some updates available for my system.

To my surprise when I happen to look at the screen (as it's taking some time
to build and I'm obviously not watching the entire time), I see this:


 * ***** WARNING *****
 * 
 * You are currently installing a version of nvidia-drivers that is
 * known not to work with a video card you have installed on your
 * system. If this is intentional, please ignore this. If it is not
 * please perform the following steps:
 * 
 * Add the following mask entry to /etc/portage/package.mask by
 * echo ">=x11-drivers/nvidia-drivers-177.0.0" >> /etc/portage/package.mask
 * 
 * Failure to perform the steps above could result in a non-working
 * X setup.
 * 
 * For more information please read:
 * http://www.nvidia.com/object/IO_32667.html
 * You must be in the video group to use the NVIDIA device
 * For more info, read the docs at
 * http://www.gentoo.org/doc/en/nvidia-guide.xml#doc_chap3_sect6
 * 
 * This ebuild installs a kernel module and X driver. Both must
 * match explicitly in their version. This means, if you restart
 * X, you most modprobe -r nvidia before starting it back up
 * 
 * To use the NVIDIA GLX, run "eselect opengl set nvidia"
 * 
 * nVidia has requested that any bug reports submitted have the
 * output of /usr/bin/nvidia-bug-report.sh included.
 * 
 * To work with compiz, you must enable the AddARGBGLXVisuals option.
 * 
 * If you are having resolution problems, try disabling DynamicTwinView.

Sure enough, X no longer works. I'm following the instructions now, but...
Don't you think the default action here should be to do nothing instead of
breaking my system?

Not impressed. Hopefully this critical message would be summarized at the end
of the build too. Kind of important. I got lucky and happened to see it...

Thanks,
Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Stroller]

On 31 Dec 2008, at 23:33, Michael P. Soulier wrote:
> ...
> Don't you think the default action here should be to do nothing  
> instead of
> breaking my system?


   "That proposal is ludicrous and completely counter to the Unix
    way of doing things."

Not my opinion, just quoting.

Stroller.

Re: [gentoo-user] nvidia warning comes a tad late

[Graham Murray]

"Michael P. Soulier" <msoulier@digitaltorque.ca> writes:

> Sure enough, X no longer works. I'm following the instructions now, but...
> Don't you think the default action here should be to do nothing instead of
> breaking my system?

I think that the default action should be that such 'breakages' should
be checked during the dependency building phase, a message displayed and
the emerge stop[0]. Then you could either mask the offending package or
issue a special flag[1] to emerge to acknowledge the 'problem' but
install/upgrade the package anyway.

[0] As with package blockers.

[1] A new flag, something like '--unsafe'

Re: [gentoo-user] nvidia warning comes a tad late

[Volker Armin Hemmann]

On Donnerstag 01 Januar 2009, Michael P. Soulier wrote:

>
> Not impressed. Hopefully this critical message would be summarized at the
> end of the build too. Kind of important. I got lucky and happened to see
> it...

it was. Also:
elog
and
elogv

the tools are there. It is your fault of not using them.

Re: [gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 515 bytes --]

On 01/01/09 Volker Armin Hemmann said:

> it was. Also:
> elog
> and
> elogv
> 
> the tools are there. It is your fault of not using them.

Great, please demonstrate how I was to know about this breakage before it
happened, and I'll change how I use the tools. 

Cheers,
Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 711 bytes --]

On 01/01/09 Graham Murray said:

> I think that the default action should be that such 'breakages' should
> be checked during the dependency building phase, a message displayed and
> the emerge stop[0]. Then you could either mask the offending package or
> issue a special flag[1] to emerge to acknowledge the 'problem' but
> install/upgrade the package anyway.
> 
> [0] As with package blockers.
> 
> [1] A new flag, something like '--unsafe'

I completely agree.

Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Volker Armin Hemmann]

On Donnerstag 01 Januar 2009, Michael P. Soulier wrote:
> On 01/01/09 Volker Armin Hemmann said:
> > it was. Also:
> > elog
> > and
> > elogv
> >
> > the tools are there. It is your fault of not using them.
>
> Great, please demonstrate how I was to know about this breakage before it
> happened, and I'll change how I use the tools.
>
> Cheers,
> Mike

after the emerge you read the messages with elogv and downgrade. No harm done.

Re: [gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 711 bytes --]

On 01/01/09 Volker Armin Hemmann said:

> after the emerge you read the messages with elogv and downgrade. No harm done.

I'll be sure to try that, thank you. However, would not avoiding a bad upgrade
in the first place be a better-behaved tool? Especially when the package in
question "knew" that it was likely incompatible?

I'm not saying that this could not be avoided with more work, I'm saying that
I shouldn't have to if the tools were better behaved.

Cheers,
Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Volker Armin Hemmann]

On Donnerstag 01 Januar 2009, Michael P. Soulier wrote:
> On 01/01/09 Volker Armin Hemmann said:
> > after the emerge you read the messages with elogv and downgrade. No harm
> > done.
>
> I'll be sure to try that, thank you. However, would not avoiding a bad
> upgrade in the first place be a better-behaved tool? Especially when the
> package in question "knew" that it was likely incompatible?
>
> I'm not saying that this could not be avoided with more work, I'm saying
> that I shouldn't have to if the tools were better behaved.
>
> Cheers,
> Mike

how should 'the tool' know what card you are using? and even if portage could 
parse lspci output - why make it slower and more easily to break if all 
breakage can be avoided by simply reading first - then upgrading? Do you 
always install the latest drivers without reading up on them first?
Nvidia's 'deprecation' strategy is a pain in the ass and they are doing it for 
a long time now. So this time it bit you. Next time it will be 6XXX card 
users, then 7XXX card users and so on. That is why you have to go to nvnews 
first and then upgrade. Not the other way round.

Re: [gentoo-user] nvidia warning comes a tad late

[Dale]

Michael P. Soulier wrote:
> So, like a good gentoo user I'm emerging some updates available for my system.
>
> To my surprise when I happen to look at the screen (as it's taking some time
> to build and I'm obviously not watching the entire time), I see this:
>
>
>  * ***** WARNING *****
>  * 
>  * You are currently installing a version of nvidia-drivers that is
>  * known not to work with a video card you have installed on your
>  * system. If this is intentional, please ignore this. If it is not
>  * please perform the following steps:
>  * 
>  * Add the following mask entry to /etc/portage/package.mask by
>  * echo ">=x11-drivers/nvidia-drivers-177.0.0" >> /etc/portage/package.mask
>  * 
>  * Failure to perform the steps above could result in a non-working
>  * X setup.
>  * 
>  * For more information please read:
>  * http://www.nvidia.com/object/IO_32667.html
>  * You must be in the video group to use the NVIDIA device
>  * For more info, read the docs at
>  * http://www.gentoo.org/doc/en/nvidia-guide.xml#doc_chap3_sect6
>  * 
>  * This ebuild installs a kernel module and X driver. Both must
>  * match explicitly in their version. This means, if you restart
>  * X, you most modprobe -r nvidia before starting it back up
>  * 
>  * To use the NVIDIA GLX, run "eselect opengl set nvidia"
>  * 
>  * nVidia has requested that any bug reports submitted have the
>  * output of /usr/bin/nvidia-bug-report.sh included.
>  * 
>  * To work with compiz, you must enable the AddARGBGLXVisuals option.
>  * 
>  * If you are having resolution problems, try disabling DynamicTwinView.
>
> Sure enough, X no longer works. I'm following the instructions now, but...
> Don't you think the default action here should be to do nothing instead of
> breaking my system?
>
> Not impressed. Hopefully this critical message would be summarized at the end
> of the build too. Kind of important. I got lucky and happened to see it...
>
> Thanks,
> Mike
>   

I just did a reinstall on my rig and it did the exact same thing.  I had
to mask the one it installed and re-emerge the older one that does
work.  Isn't there some way for it to pick the right one?  After all, it
new it was the WRONG one it was installing.  Looks to me like it could
pick the right one.

Dale

:-)  :-) 

[gentoo-user] Re: nvidia warning comes a tad late

[»Q«]

In <200901010423.25783.volker.armin.hemmann@tu-clausthal.de>,
Volker Armin Hemmann <volker.armin.hemmann@tu-clausthal.de> wrote:

> That is why you have to go to nvnews first and then upgrade.

Where is nvnews?  

I've been going to <http://www.nvidia.com/object/unix.html>, selecting
the driver I'm thinking of upgrading to, and checking its compatibility
list.

-- 
»Q«
     Kleeneness is next to Gödelness.

Re: [gentoo-user] nvidia warning comes a tad late

[Alan McKinnon]

On Thursday 01 January 2009 02:25:10 Stroller wrote:
> On 31 Dec 2008, at 23:33, Michael P. Soulier wrote:
> > ...
> > Don't you think the default action here should be to do nothing
> > instead of
> > breaking my system?
>
>    "That proposal is ludicrous and completely counter to the Unix
>     way of doing things."
>
> Not my opinion, just quoting.

nice one :-)

The Unix way is to do what the user told it to do, no more and no less.

If you tell the system to install a driver, ignore the prompt or even 
type "y", why are users constantly surprised when the system does exactly 
what they told it to do? What's the computer supposed to say?

"Um, no my china, look here: I don't think that's a smart move. I don't care 
what you asked, I'm just not going to do it. Eat dust, sucker"

That's how Windows works.

On Unix, if you break it you get to keep both pieces.



-- 
alan dot mckinnon at gmail dot com

Re: [gentoo-user] nvidia warning comes a tad late

[Alan McKinnon]

On Thursday 01 January 2009 11:02:23 Dale wrote:
> I just did a reinstall on my rig and it did the exact same thing.  I had
> to mask the one it installed and re-emerge the older one that does
> work.  Isn't there some way for it to pick the right one?  After all, it
> new it was the WRONG one it was installing.  Looks to me like it could
> pick the right one.

The software does not have the slightest vaguest foggiest concept of what the 
RIGHT and the WRONG drivers are. That's a human being's conclusion.

It therefore cannot decide.
The devs therefore correctly decided to not even try and decide.

Unix-like systems demand that the user actually has a clue, is more than a 
mere automatonic moron, can and does read information and can and does really 
make decisions. And is prepared to live with the results.

Some Unix people try to get all politically correct and hide this fundamental 
fact, but that is just plain wrong. It will never work any other way than how 
it is working right now.

Users that are not prepared to actually think about what they are doing should 
switch back to Windows. That system specializes in treating their customers 
like complete idiots.

-- 
alan dot mckinnon at gmail dot com

Re: [gentoo-user] nvidia warning comes a tad late

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1236 bytes --]

On Thu, 1 Jan 2009 12:27:48 +0200, Alan McKinnon wrote:

> > > Don't you think the default action here should be to do nothing
> > > instead of
> > > breaking my system?  
> >
> >    "That proposal is ludicrous and completely counter to the Unix
> >     way of doing things."
> >
> > Not my opinion, just quoting.  
> 
> nice one :-)
> 
> The Unix way is to do what the user told it to do, no more and no less.
> 
> If you tell the system to install a driver, ignore the prompt or even 
> type "y", why are users constantly surprised when the system does
> exactly what they told it to do? What's the computer supposed to say?

Except in this case, portage knew the action was risky but issued the
warning after the event "you really shouldn't have done that", like a
typical smartarse with20:20 hindsight.

There are numerous examples of ebuilds that stop if an upgrade is risky,
postfix is one such, and provide the user with the an option to
carry on if they choose, usually be setting an environment variable.

I really don't see the point in an ebuild making this sort of test and
then continuing to install anyway.


-- 
Neil Bothwick

I couldn't repair your brakes, so I made your horn louder.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Philip Webb]

090101 Neil Bothwick wrote:
> On Thu, 1 Jan 2009 12:27:48 +0200, Alan McKinnon wrote:
>>> Don't you think the default action here should be to do nothing
>>> instead of breaking my system?  
>> If you tell the system to install a driver, ignore the prompt
>> or even type "y", why are users constantly surprised
>> when the system does exactly what they told it to do?
> Except in this case, portage knew the action was risky
> but issued the warning after the event
> "you really shouldn't have done that", like a typical smartarse.
> There are numerous examples of ebuilds that stop if an upgrade is risky,
> postfix is one such, and provide the user with the an option to carry on
> if they choose, usually be setting an environment variable.
> I really don't see the point in an ebuild making this sort of test
> and then continuing to install anyway.

I agree.  I ran into this on my back-up box which has an older card,
but as I never do 'emerge world' without '-pv', I saw it in time
& aborted via '^c'.  I've now made a prominent note in my pkg list
for that machine not to try to upgrade the Nvidia driver.

Portage knows that what is proposed is going to break the user's system,
so it should refuse to do it.  It's like "Package A blocks package B",
which causes the emerge to stop till the user acts more sensibly.

-- 
========================,,============================================
SUPPORT     ___________//___,   Philip Webb
ELECTRIC   /] [] [] [] [] []|   Cities Centre, University of Toronto
TRANSIT    `-O----------O---'   purslowatchassdotutorontodotca

Re: [gentoo-user] nvidia warning comes a tad late

[Dale]

Alan McKinnon wrote:
> On Thursday 01 January 2009 11:02:23 Dale wrote:
>   
>> I just did a reinstall on my rig and it did the exact same thing.  I had
>> to mask the one it installed and re-emerge the older one that does
>> work.  Isn't there some way for it to pick the right one?  After all, it
>> new it was the WRONG one it was installing.  Looks to me like it could
>> pick the right one.
>>     
>
> The software does not have the slightest vaguest foggiest concept of what the 
> RIGHT and the WRONG drivers are. That's a human being's conclusion.
>
> It therefore cannot decide.
> The devs therefore correctly decided to not even try and decide.
>
> Unix-like systems demand that the user actually has a clue, is more than a 
> mere automatonic moron, can and does read information and can and does really 
> make decisions. And is prepared to live with the results.
>
> Some Unix people try to get all politically correct and hide this fundamental 
> fact, but that is just plain wrong. It will never work any other way than how 
> it is working right now.
>
> Users that are not prepared to actually think about what they are doing should 
> switch back to Windows. That system specializes in treating their customers 
> like complete idiots.
>
>   

Not disputing what you say but if it doesn't know what card we are
using, why does it warn us that it is not compatible?  Exact same thing
happened to me a couple weeks ago and it has not happened before that,
that I can recall anyway.

Dale

:-)  :-) 

Re: [gentoo-user] nvidia warning comes a tad late

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 596 bytes --]

On Thu, 1 Jan 2009 05:54:33 -0500, Philip Webb wrote:

> Portage knows that what is proposed is going to break the user's system,
> so it should refuse to do it.  It's like "Package A blocks package B",
> which causes the emerge to stop till the user acts more sensibly.

This is different in that the problem is not detected until the emerge
starts, but portage could skip this package and carry on with the rest,
issuing an elog message explaining what happened and how to force an
install if that's what you really want.


-- 
Neil Bothwick

I cna ytpe 300 wrods pre mniuet!!!

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[b.n.]

Volker Armin Hemmann ha scritto:
> On Donnerstag 01 Januar 2009, Michael P. Soulier wrote:
>> On 01/01/09 Volker Armin Hemmann said:
>>> after the emerge you read the messages with elogv and downgrade. No harm
>>> done.
>> I'll be sure to try that, thank you. However, would not avoiding a bad
>> upgrade in the first place be a better-behaved tool? Especially when the
>> package in question "knew" that it was likely incompatible?
>>
>> I'm not saying that this could not be avoided with more work, I'm saying
>> that I shouldn't have to if the tools were better behaved.
>>
>> Cheers,
>> Mike
> 
> how should 'the tool' know what card you are using? 

The tool knew -in fact it told him of the breakage , *after* doing it.

> and even if portage could 
> parse lspci output - why make it slower and more easily to break if all 
> breakage can be avoided by simply reading first - then upgrading? 

If you don't know there's something to read...

> Do you 
> always install the latest drivers without reading up on them first?

Usually, yes. Could be my fault, but am I expected to read technical
docs everytime I update a package?
Anyway, the system *knows* that there's a problem, so your point is
moot. The only thing we're asking is to warn and stop *before* and not
*after*.

> Nvidia's 'deprecation' strategy is a pain in the ass and they are doing it for 
> a long time now. So this time it bit you. Next time it will be 6XXX card 
> users, then 7XXX card users and so on. That is why you have to go to nvnews 
> first and then upgrade. Not the other way round.

Thanks for advice.

m.

Re: [gentoo-user] nvidia warning comes a tad late

[Matt Causey]

I am  total Gentoo newb :D  but it seems kind of fundamental to the
concept of this distribution that its users are going to make
themselves aware of the details of system updates.  Short of reading
ridiculous amounts of doco...folks should be reading the output of the
emerge commands to learn about edge cases like this one.

In the short few days I've been using Gentoo, there have been several
occasions where had I not read that output, my system would have been
'broken' on next reboot.  At the very least there were additional
steps needed for me to install that package I tried to emerge (missing
USE flags, requests to rebuild other packages, external data
downloads, etc.).

Personally, I rather like this approach.  The folks maintaining the
builds take the time to identify these edge cases, which makes the
portage text output quite helpful.

--
Matt

On Thu, Jan 1, 2009 at 1:09 PM, b.n. <brullonulla@gmail.com> wrote:
> Volker Armin Hemmann ha scritto:
>> On Donnerstag 01 Januar 2009, Michael P. Soulier wrote:
>>> On 01/01/09 Volker Armin Hemmann said:
>>>> after the emerge you read the messages with elogv and downgrade. No harm
>>>> done.
>>> I'll be sure to try that, thank you. However, would not avoiding a bad
>>> upgrade in the first place be a better-behaved tool? Especially when the
>>> package in question "knew" that it was likely incompatible?
>>>
>>> I'm not saying that this could not be avoided with more work, I'm saying
>>> that I shouldn't have to if the tools were better behaved.
>>>
>>> Cheers,
>>> Mike
>>
>> how should 'the tool' know what card you are using?
>
> The tool knew -in fact it told him of the breakage , *after* doing it.
>
>> and even if portage could
>> parse lspci output - why make it slower and more easily to break if all
>> breakage can be avoided by simply reading first - then upgrading?
>
> If you don't know there's something to read...
>
>> Do you
>> always install the latest drivers without reading up on them first?
>
> Usually, yes. Could be my fault, but am I expected to read technical
> docs everytime I update a package?
> Anyway, the system *knows* that there's a problem, so your point is
> moot. The only thing we're asking is to warn and stop *before* and not
> *after*.
>
>> Nvidia's 'deprecation' strategy is a pain in the ass and they are doing it for
>> a long time now. So this time it bit you. Next time it will be 6XXX card
>> users, then 7XXX card users and so on. That is why you have to go to nvnews
>> first and then upgrade. Not the other way round.
>
> Thanks for advice.
>
> m.
>
>
>
>
>

[gentoo-user] Re: nvidia warning comes a tad late

[Grant Edwards]

On 2009-01-01, Matt Causey <matt.causey@gmail.com> wrote:

> I am total Gentoo newb :D but it seems kind of fundamental to
> the concept of this distribution that its users are going to
> make themselves aware of the details of system updates.  Short
> of reading ridiculous amounts of doco...folks should be
> reading the output of the emerge commands to learn about edge
> cases like this one.

There are plenty of ebuilds that when they know they are going
to break the system will abort with a warning to the user how
to either prevent the breakage or how to force the install.  I
don't see any reason why the nvidia ebuild should go ahead
and break the system and then tell you about it afterwards.

Why not tell you about how the update will break your system
and then _not_ doing the update?

> Personally, I rather like this approach.  The folks
> maintaining the builds take the time to identify these edge
> cases, which makes the portage text output quite helpful.

It would be even more helpful if the ebuild _doesn't_ break
your system.

-- 
Grant

Re: [gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 1852 bytes --]

On 01/01/09 Alan McKinnon said:

> The software does not have the slightest vaguest foggiest concept of what the 
> RIGHT and the WRONG drivers are. That's a human being's conclusion.

Apparently it did, hence the warning.

> It therefore cannot decide.

It did decide. It decided to continue.

> The devs therefore correctly decided to not even try and decide.
> 
> Unix-like systems demand that the user actually has a clue, is more than a 
> mere automatonic moron, can and does read information and can and does really 
> make decisions. And is prepared to live with the results.

Orthogonal to the discussion. You are blaming users for laziness in the system
that could have made it easier to notice a potential problem.

> Some Unix people try to get all politically correct and hide this fundamental 
> fact, but that is just plain wrong. It will never work any other way than how 
> it is working right now.

Justification by tradition won't help anyone here. I see nothing in this post
but inflammatory, flawed logic.

> Users that are not prepared to actually think about what they are doing should 
> switch back to Windows. That system specializes in treating their customers 
> like complete idiots.

Like this statement. 

I see many posts like this but few suggestions as to how the problem could
have been avoided ahead of time. I saw one suggestion of how to roll the
driver back after the fact, which I did, after it was already broken.

Does anyone have any rational arguments to support the system not stopping due
to the warning, or is this all I can expect?

Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 557 bytes --]

On 01/01/09 Alan McKinnon said:

> nice one :-)
> 
> The Unix way is to do what the user told it to do, no more and no less.
> 
> If you tell the system to install a driver, ignore the prompt or even 

Ignore what prompt? There was no prompt, a prompt requiring feedback is in
fact, exactly what I am looking for.

Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

[gentoo-user] Re: nvidia warning comes a tad late

[Nikos Chantziaras]

Michael P. Soulier wrote:
> [...]
> I see many posts like this but few suggestions as to how the problem could
> have been avoided ahead of time.

You can open a bug about it and suggest something.

Re: [gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 949 bytes --]

On 01/01/09 Neil Bothwick said:

> This is different in that the problem is not detected until the emerge
> starts, but portage could skip this package and carry on with the rest,
> issuing an elog message explaining what happened and how to force an
> install if that's what you really want.

Yes, that would have been helpful. The message in fact was very helpful in
showing me how to fix the problem, and I am thankful that the effort was taken
as Gentoo is still a little new to me (I come from Debian/RedHat land mostly).
I'm not against the warning, as the subject of this thread states, it just
came a little late. :)

I like Gentoo, but I find it in the wrong in this particular case.

Cheers,
Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] Re: nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 373 bytes --]

On 01/01/09 Nikos Chantziaras said:

> You can open a bug about it and suggest something.

I did yesterday when it happened. 

Thanks,
Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Dirk Heinrichs]

[-- Attachment #1: Type: text/plain, Size: 332 bytes --]

Am Donnerstag, 1. Januar 2009 00:33:27 schrieb Michael P. Soulier:

> Don't you think the default action here should be to do nothing instead of
> breaking my system?

What we think here is irrelevant. You should file a bug and see what the devs 
think. We can then express what we think by voting for it.

Bye...

	Dirk

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 566 bytes --]

On Thu, 1 Jan 2009 11:26:27 -0500, Michael P. Soulier wrote:

> Ignore what prompt? There was no prompt, a prompt requiring feedback is
> in fact, exactly what I am looking for.

That would be wrong. Emerge is supposed to run non-interactively, apart
from a prompt at the start of the process when using --ask. A world
update can take many hours and is often run overnight, imagine your
frustration the next morning when you see it is asking if you want to
proceed on package 3/184.


-- 
Neil Bothwick

Pedestrians come in two types: Quick or Dead.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Dale]

Neil Bothwick wrote:
> On Thu, 1 Jan 2009 11:26:27 -0500, Michael P. Soulier wrote:
>
>   
>> Ignore what prompt? There was no prompt, a prompt requiring feedback is
>> in fact, exactly what I am looking for.
>>     
>
> That would be wrong. Emerge is supposed to run non-interactively, apart
> from a prompt at the start of the process when using --ask. A world
> update can take many hours and is often run overnight, imagine your
> frustration the next morning when you see it is asking if you want to
> proceed on package 3/184.
>
>
>   

Only thing worse is a system that completed the emerge but don't work. 
I have one word:  Great!  That is a bit sarcastic to.  I can think of a
couple other words but you may get the idea.  The others are. . . dirty.

Dale

:-)  :-) 

Re: [gentoo-user] Re: nvidia warning comes a tad late

[Volker Armin Hemmann]

On Donnerstag 01 Januar 2009, »Q« wrote:
> In <200901010423.25783.volker.armin.hemmann@tu-clausthal.de>,
>
> Volker Armin Hemmann <volker.armin.hemmann@tu-clausthal.de> wrote:
> > That is why you have to go to nvnews first and then upgrade.
>
> Where is nvnews?

http://www.nvnews.net/vbulletin/forumdisplay.php?s=&forumid=14

Re: [gentoo-user] nvidia warning comes a tad late

[Volker Armin Hemmann]

On Donnerstag 01 Januar 2009, Neil Bothwick wrote:
> On Thu, 1 Jan 2009 12:27:48 +0200, Alan McKinnon wrote:
> > > > Don't you think the default action here should be to do nothing
> > > > instead of
> > > > breaking my system?
> > >
> > >    "That proposal is ludicrous and completely counter to the Unix
> > >     way of doing things."
> > >
> > > Not my opinion, just quoting.
> >
> > nice one :-)
> >
> > The Unix way is to do what the user told it to do, no more and no less.
> >
> > If you tell the system to install a driver, ignore the prompt or even
> > type "y", why are users constantly surprised when the system does
> > exactly what they told it to do? What's the computer supposed to say?
>
> Except in this case, portage knew the action was risky but issued the
> warning after the event "you really shouldn't have done that", like a
> typical smartarse with20:20 hindsight.
>
> There are numerous examples of ebuilds that stop if an upgrade is risky,
> postfix is one such, and provide the user with the an option to
> carry on if they choose, usually be setting an environment variable.
>
> I really don't see the point in an ebuild making this sort of test and
> then continuing to install anyway.

but as long as X is not restarted, the upgrade doesn't break anything. You 
come back, you read the elogs, you downgrade the drivers and everything is 
fine and dandy.

Re: [gentoo-user] nvidia warning comes a tad late

[Volker Armin Hemmann]

On Donnerstag 01 Januar 2009, Michael P. Soulier wrote:
> On 01/01/09 Alan McKinnon said:
> > The software does not have the slightest vaguest foggiest concept of what
> > the RIGHT and the WRONG drivers are. That's a human being's conclusion.
>
> Apparently it did, hence the warning.

the ebuild warned you. Portage and ebuilds are different things. And portage 
has to assume that you know what you are doing.

>
> > It therefore cannot decide.
>
> It did decide. It decided to continue.

because it SUCKS when a world update breaks somewhere along 25 of 223. People 
don't want portage to stop.

>
> > The devs therefore correctly decided to not even try and decide.
> >
> > Unix-like systems demand that the user actually has a clue, is more than
> > a mere automatonic moron, can and does read information and can and does
> > really make decisions. And is prepared to live with the results.
>
> Orthogonal to the discussion. You are blaming users for laziness in the
> system that could have made it easier to notice a potential problem.

the user is the only one to blame - if you restart X or your system before 
reading the elogs, it is your own fault if something breaks. A running 
service, like X, ssh, apache, isn't influenced by any update until you restart 
it.

So a user who didn't read up before updating and then doesn't read after it 
too deserves what he get.

>
> > Some Unix people try to get all politically correct and hide this
> > fundamental fact, but that is just plain wrong. It will never work any
> > other way than how it is working right now.
>
> Justification by tradition won't help anyone here. I see nothing in this
> post but inflammatory, flawed logic.

no, he is right. Linux is not Windows. There are some people who want to turn 
linux into windows. These people should buy a mac.

>
> > Users that are not prepared to actually think about what they are doing
> > should switch back to Windows. That system specializes in treating their
> > customers like complete idiots.
>
> Like this statement.
>
> I see many posts like this but few suggestions as to how the problem could
> have been avoided ahead of time. I saw one suggestion of how to roll the
> driver back after the fact, which I did, after it was already broken.
>
> Does anyone have any rational arguments to support the system not stopping
> due to the warning, or is this all I can expect?

BECAUSE STOPPING IS EVIL! PORTAGE IS NON INTERACTIVE! People want to start an 
update then go away or sleep. I think Neil already told you that.

Re: [gentoo-user] nvidia warning comes a tad late

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 543 bytes --]

On Thu, 1 Jan 2009 18:42:23 +0100, Volker Armin Hemmann wrote:

> BECAUSE STOPPING IS EVIL! PORTAGE IS NON INTERACTIVE! People want to
> start an update then go away or sleep. I think Neil already told you
> that.

Yes I did. But I also stated that I believe portage should skip the
package when this situation occurs, unless you have explicitly told it to
proceed with the potentially broken version.


-- 
Neil Bothwick

Why marry a virgin? If she wasn't good enough for the rest of them, then
she isn't good enough for you.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 831 bytes --]

On Thu, 1 Jan 2009 18:34:36 +0100, Volker Armin Hemmann wrote:

> but as long as X is not restarted, the upgrade doesn't break anything.
> You come back, you read the elogs, you downgrade the drivers and
> everything is fine and dandy.

Except you've wasted time and resources compiling the broken version of
the software and then recompiling the version you already had. If the
ebuild, in fact it's the nvidia.eclass, can detect that proceeding will
cause breakage, why proceed?

Then there's the case where an update to another package now prevents the
old one from compiling. It shouldn't happen, but it does, so why risk all
those disadvantages when portage could use its --keep-going code to
restart the emerge with the net package?


-- 
Neil Bothwick

God is real, unless specifically declared integer.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 2448 bytes --]

On 01/01/09 Volker Armin Hemmann said:

> the ebuild warned you. Portage and ebuilds are different things. And portage 
> has to assume that you know what you are doing.

Sure, the issue is that it warned me too late.

> because it SUCKS when a world update breaks somewhere along 25 of 223. People 
> don't want portage to stop.

Perhaps then all such checks should be done at the beginning of running
portage, instead of at the beginning of the individual builds. Debian does
this, running all pre-scripts before actually installing the packages. There
are more than two options here.

> the user is the only one to blame - if you restart X or your system before 
> reading the elogs, it is your own fault if something breaks. A running 
> service, like X, ssh, apache, isn't influenced by any update until you restart 
> it.

No, untrue. Running services with loadable modules such as apache can easily
be disasterously influenced by underlying changes while they are running. I've
seen it many times.

> So a user who didn't read up before updating and then doesn't read after it 
> too deserves what he get.

I was upgrading on the order of 20 packages. Thank goodness I didn't deploy
Gentoo in an enterprise environment and only broke the single machine. Your
philosophy seems to put an undue amount of work on the administrator. Exactly
how many websites should I be checking before I follow the simplistic
instructions in the Gentoo handbook that tell me to just "emerge --update
world"? I followed the instructions found here

http://www.gentoo.org/doc/en/handbook/handbook-x86.xml?part=2&chap=1#doc_chap3

> no, he is right. Linux is not Windows. There are some people who want to turn 
> linux into windows. These people should buy a mac.

No argument here, although I don't see how we've gotten on this side-topic of
how Linux is not Windows. I never once asked for that.

> BECAUSE STOPPING IS EVIL! PORTAGE IS NON INTERACTIVE! People want to start an 
> update then go away or sleep. I think Neil already told you that.

Which is why it's important to stop up front, not an hour into the process. Or
don't stop at all, but skip the one ebuild. 

Cheers,
Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 737 bytes --]

On 01/01/09 Neil Bothwick said:

> That would be wrong. Emerge is supposed to run non-interactively, apart
> from a prompt at the start of the process when using --ask. A world
> update can take many hours and is often run overnight, imagine your
> frustration the next morning when you see it is asking if you want to
> proceed on package 3/184.

Agreed. Skipping seems the easiest-to-implement option, as likely running all
sanity checks beforehand would likely take an architectural change.

Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Michael P. Soulier]

[-- Attachment #1: Type: text/plain, Size: 1014 bytes --]

On 01/01/09 Volker Armin Hemmann said:

> but as long as X is not restarted, the upgrade doesn't break anything. You 
> come back, you read the elogs, you downgrade the drivers and everything is 
> fine and dandy.

As long as X doesn't dynamically load a now binary-incompatible module and
segfault. X does load modules on demand from time to time, does it not? Then
of course there's the issue of power failures, my UPS only lasts for about
five minutes and we've had some wicked winter storms lately.

On another topic I'm assuming that this technique is inappropriate for
managing large numbers of workstations or servers. I assume you'd patch one
sacrificial box, and then use a completely different mechanism to push those
changes out to your managed machines.

Mike
-- 
Michael P. Soulier <msoulier@digitaltorque.ca>
"Any intelligent fool can make things bigger and more complex... It takes a
touch of genius - and a lot of courage to move in the opposite direction."
--Albert Einstein

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] nvidia warning comes a tad late

[Dale]

Neil Bothwick wrote:
> On Thu, 1 Jan 2009 18:34:36 +0100, Volker Armin Hemmann wrote:
>
>   
>> but as long as X is not restarted, the upgrade doesn't break anything.
>> You come back, you read the elogs, you downgrade the drivers and
>> everything is fine and dandy.
>>     
>
> Except you've wasted time and resources compiling the broken version of
> the software and then recompiling the version you already had. If the
> ebuild, in fact it's the nvidia.eclass, can detect that proceeding will
> cause breakage, why proceed?
>
> Then there's the case where an update to another package now prevents the
> old one from compiling. It shouldn't happen, but it does, so why risk all
> those disadvantages when portage could use its --keep-going code to
> restart the emerge with the net package?
>
>
>   

I wonder if the same would be said about something like baselayout or
some other system package that just can't be . . . screwed up?  If a new
udev would break my system and it knew it, then updated it anyway, it
wouldn't be a inconvenience at that point.  I would be pissed because I
only have one system and no way to search for a fix either.  I would be
putting a new meaning to shooting in the dark.

Dale

:-)   :-)

Re: [gentoo-user] nvidia warning comes a tad late

[Volker Armin Hemmann]

On Donnerstag 01 Januar 2009, Michael P. Soulier wrote:
> On 01/01/09 Volker Armin Hemmann said:
> > but as long as X is not restarted, the upgrade doesn't break anything.
> > You come back, you read the elogs, you downgrade the drivers and
> > everything is fine and dandy.
>
> As long as X doesn't dynamically load a now binary-incompatible module and
> segfault. X does load modules on demand from time to time, does it not?

nope. X loads everything on startup.

[gentoo-user] Re: nvidia warning comes a tad late

[»Q«]

On Thu, 1 Jan 2009 13:28:55 -0500
"Michael P. Soulier" <msoulier@digitaltorque.ca> wrote:

> Your philosophy seems to put an undue amount of work on the
> administrator.

I guess I'm in the camp that thinks the administrator should know what
modules are needed for the hardware, and portage should keep working as
it does now.

ISTM the fundamental cause of the problem is with nVidia.  Their
different series of drivers support different hardware, but instead of
distinguishing them by different package names, they only use version
numbers.  It looks like they now offer four different series,
supporting four different hardware sets (with some overlap of the sets).

IMO the best solution would be to regard the four series as four
distinct software products and give them different names.  So, e.g., if
you had installed x11-drivers/nvidia-drivers173-173.14.14, emerge -u
wouldn't install x11-drivers/nvidia-drivers177-177.82.  And people
like me, whose hardware would be supported by both packages, could
just choose which one they wanted (without having to mask anything),
which doesn't seem like too much of a burden.

Or I guess slotting could work also, but probably create
collision headaches for maintainers. 

-- 
»Q«
     Kleeneness is next to Gödelness.

Re: [gentoo-user] nvidia warning comes a tad late

[Sascha Hlusiak]

[-- Attachment #1: Type: text/plain, Size: 897 bytes --]

Am Thursday 01 January 2009 17:54:12 schrieb Neil Bothwick:
> On Thu, 1 Jan 2009 11:26:27 -0500, Michael P. Soulier wrote:
> > Ignore what prompt? There was no prompt, a prompt requiring feedback is
> > in fact, exactly what I am looking for.
>
> That would be wrong. Emerge is supposed to run non-interactively, apart
> from a prompt at the start of the process when using --ask. A world
> update can take many hours and is often run overnight, imagine your
> frustration the next morning when you see it is asking if you want to
> proceed on package 3/184.

Well, these days ebuilds can be marked as interactive, showing a yellow I in 
emerge -pv. That's what for example "doom3-demo" does with it's license, so 
if people run an emerge -uDN world over night, ignoring those flags, it's 
their fault.

I too would like to see as few interactive ebuilds as possible.

- Sascha

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-user] Re: nvidia warning comes a tad late

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 326 bytes --]

On Thu, 1 Jan 2009 14:14:16 -0600, »Q« wrote:

> I guess I'm in the camp that thinks the administrator should know what
> modules are needed for the hardware, and portage should keep working as
> it does now.

Then why the test and warning?


-- 
Neil Bothwick

God is real, unless specifically declared integer.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

[gentoo-user] Re: nvidia warning comes a tad late

[»Q«]

In <20090101213152.77d30b94@krikkit>,
Neil Bothwick <neil@digimed.co.uk> wrote:

> On Thu, 1 Jan 2009 14:14:16 -0600, »Q« wrote:
> 
> > I guess I'm in the camp that thinks the administrator should know
> > what modules are needed for the hardware, and portage should keep
> > working as it does now.  
> 
> Then why the test and warning?

I haven't advocated a test and warning.  But why not?

-- 
»Q«
     Kleeneness is next to Gödelness.

Re: [gentoo-user] nvidia warning comes a tad late

[Stroller]

On 1 Jan 2009, at 03:23, Volker Armin Hemmann wrote:
> ...
> [assuming] portage could
> parse lspci output - why make it slower and more easily to break if  
> all
> breakage can be avoided by simply reading first - then upgrading?

We have computers to make our lives simpler & easier. If a computer  
can automatically detect breakage & avoid it, then it saves the user  
reading documentation for many packages.

Stroller.

Re: [gentoo-user] nvidia warning comes a tad late

[Joshua Murphy]

On Wed, Dec 31, 2008 at 7:53 PM, Graham Murray <graham@gmurray.org.uk> wrote:
> "Michael P. Soulier" <msoulier@digitaltorque.ca> writes:
>
>> Sure enough, X no longer works. I'm following the instructions now, but...
>> Don't you think the default action here should be to do nothing instead of
>> breaking my system?
>
> I think that the default action should be that such 'breakages' should
> be checked during the dependency building phase, a message displayed and
> the emerge stop[0]. Then you could either mask the offending package or
> issue a special flag[1] to emerge to acknowledge the 'problem' but
> install/upgrade the package anyway.
>
> [0] As with package blockers.
>
> [1] A new flag, something like '--unsafe'

"[1]" there isn't as new as you might think, though it's a variable
rather than a flag... I quote from the Busybox ebuild:

"set env VERY_BRAVE_OR_VERY_DUMB=yes if this is realy what you want."
"silly options will destroy your system"

For any that haven't played with emerging Busybox to ROOT=/ and
USE=make-symlinks, the text above is an excerpt from the message when
the emerge chastises you ad calls 'die'.

Incidentally... I'm going to go bug that typo (realy)... yay for
Firefox's built in spell checking.

-- 
Poison [BLX]
Joshua M. Murphy

Re: [gentoo-user] Re: nvidia warning comes a tad late

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 837 bytes --]

On Thu, 1 Jan 2009 16:39:53 -0600, »Q« wrote:

> > > I guess I'm in the camp that thinks the administrator should know
> > > what modules are needed for the hardware, and portage should keep
> > > working as it does now.    
> > 
> > Then why the test and warning?  
> 
> I haven't advocated a test and warning.  But why not?

That's the point of this thread, the ebuild does perform a test before
installation, but goes ahead straight after the warning. Unless you are
watching the screen at that exact moment, you won't know your system was
broken until you read the post-emerge messages - and they won't appear in
the terminal if a subsequent, unconnected, emerge fails. While it's
better than no warning at all, it's completely arse-about-face.


-- 
Neil Bothwick

Talk is cheap because supply exceeds demand.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-user] Re: nvidia warning comes a tad late

[Dirk Heinrichs]

[-- Attachment #1: Type: text/plain, Size: 240 bytes --]

Am Donnerstag, 1. Januar 2009 23:39:53 schrieb »Q«:

> > Then why the test and warning?
>
> I haven't advocated a test and warning.  But why not?

There _is_ a test and warning. See very first mail in this thread.

Bye...

	Dirk

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[gentoo-user] Re: nvidia warning comes a tad late

[»Q«]

On Fri, 2 Jan 2009 11:26:20 +0000
Neil Bothwick <neil@digimed.co.uk> wrote:

> On Thu, 1 Jan 2009 16:39:53 -0600, »Q« wrote:
> 
> > > > I guess I'm in the camp that thinks the administrator should
> > > > know what modules are needed for the hardware, and portage
> > > > should keep working as it does now.    
> > > 
> > > Then why the test and warning?  
> > 
> > I haven't advocated a test and warning.  But why not?
> 
> That's the point of this thread, the ebuild does perform a test before
> installation, but goes ahead straight after the warning. 

AFAIAC, the post-install log is exactly where the message belongs --
that's where I'd look if I'd broken my system.  The fact that I don't
think portage should prevent people from installing stuff doesn't mean
I think there shouldn't be any information about what they've just
installed.

But you snipped without comment what I think was a better idea, just
making the 177.x series no longer be an upgrade to the 173.x series.

-- 
»Q«
     Kleeneness is next to Gödelness.

[gentoo-user] Re: nvidia warning comes a tad late

[»Q«]

On Fri, 2 Jan 2009 12:35:15 +0100
Dirk Heinrichs <dirk.heinrichs@online.de> wrote:

> Am Donnerstag, 1. Januar 2009 23:39:53 schrieb »Q«:
> 
> > > Then why the test and warning?  
> >
> > I haven't advocated a test and warning.  But why not?  
> 
> There _is_ a test and warning. See very first mail in this thread.

I've followed it all, and I know there's a test and warning.  Just
wasn't sure why Neil was asking *me* about why there's a warning.

-- 
»Q«
     Kleeneness is next to Gödelness.

Re: [gentoo-user] Re: nvidia warning comes a tad late

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1363 bytes --]

On Fri, 2 Jan 2009 09:09:23 -0600, »Q« wrote:

> > That's the point of this thread, the ebuild does perform a test before
> > installation, but goes ahead straight after the warning.   
> 
> AFAIAC, the post-install log is exactly where the message belongs --
> that's where I'd look if I'd broken my system.

Would it be better if your system wasn't broken?

> The fact that I don't
> think portage should prevent people from installing stuff doesn't mean
> I think there shouldn't be any information about what they've just
> installed.

There is another option,and it's already used in other ebuilds. Warn and
abort emerging that package unless the user has specified that it
should be installed.

> But you snipped without comment what I think was a better idea, just
> making the 177.x series no longer be an upgrade to the 173.x series.

Making different packages is one idea, but will still cause problems in
the future. The latest package,whatever you name it, would be the
correct one for7/8/9xxx cards,but at some time it would drop support for
7xxx cards. Maybe a better option would be a make.conf variable, like
NVIDIA_VIDEO_CARD, that ebuilds would respect in deciding which versionto
use.


-- 
Neil Bothwick

The trouble with doing something right the first time is that nobody
appreciates how difficult it was.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

[gentoo-user] Re: nvidia warning comes a tad late

[»Q«]

In <20090102224554.57ea4a64@krikkit>,
Neil Bothwick <neil@digimed.co.uk> wrote:

> On Fri, 2 Jan 2009 09:09:23 -0600, »Q« wrote:
> 
> > > That's the point of this thread, the ebuild does perform a test
> > > before installation, but goes ahead straight after the warning.   
> > 
> > AFAIAC, the post-install log is exactly where the message belongs --
> > that's where I'd look if I'd broken my system.
> 
> Would it be better if your system wasn't broken?

Yes, but I continue not to believe that it should be portage's job to
prevent me from installing things that break my system.

> > The fact that I don't
> > think portage should prevent people from installing stuff doesn't
> > mean I think there shouldn't be any information about what they've
> > just installed.
> 
> There is another option,and it's already used in other ebuilds. Warn
> and abort emerging that package unless the user has specified that it
> should be installed.

Is it only aborted if the command was --update world, or would it also
be aborted if the problem package was part of some other set?  (I hope
the question makes sense -- I haven't followed all the newish stuff
about sets of packages.)

> > But you snipped without comment what I think was a better idea, just
> > making the 177.x series no longer be an upgrade to the 173.x series.
> 
> Making different packages is one idea, but will still cause problems
> in the future. The latest package,whatever you name it, would be the
> correct one for7/8/9xxx cards,but at some time it would drop support
> for 7xxx cards. 

Don't nVidia give it a new major version number when they drop support,
so that the latest new package at that time would get a new name?  If
they *do* drop support even within a major version, my idea wouldn't
stand a chance of working well.

> Maybe a better option would be a make.conf variable, like
> NVIDIA_VIDEO_CARD, that ebuilds would respect in deciding which
> versionto use.

I like that idea better than mine.

-- 
»Q«
     Kleeneness is next to Gödelness.

[gentoo-user] Re: nvidia warning comes a tad late

[Grant Edwards]

On 2009-01-02, ?Q? <boxcars@gmx.net> wrote:
> In <20090102224554.57ea4a64@krikkit>,
> Neil Bothwick <neil@digimed.co.uk> wrote:
>
>> On Fri, 2 Jan 2009 09:09:23 -0600, ?Q? wrote:
>> 
>> > > That's the point of this thread, the ebuild does perform a test
>> > > before installation, but goes ahead straight after the warning.   
>> > 
>> > AFAIAC, the post-install log is exactly where the message belongs --
>> > that's where I'd look if I'd broken my system.
>> 
>> Would it be better if your system wasn't broken?
>
> Yes, but I continue not to believe that it should be portage's job to
> prevent me from installing things that break my system.

You must be pretty unhappy with Gentoo, because portage seems
to go to a great deal of effort to avoid breaking things (what
with all that dependancy stuff it does).  Several times a month
it refuses to update because of blockages alone.

-- 
Grant

Re: [gentoo-user] Re: nvidia warning comes a tad late

[Dale]

Grant Edwards wrote:
> On 2009-01-02, ?Q? <boxcars@gmx.net> wrote:
>   
>> In <20090102224554.57ea4a64@krikkit>,
>> Neil Bothwick <neil@digimed.co.uk> wrote:
>>
>>     
>>> On Fri, 2 Jan 2009 09:09:23 -0600, ?Q? wrote:
>>>
>>>       
>>>>> That's the point of this thread, the ebuild does perform a test
>>>>> before installation, but goes ahead straight after the warning.   
>>>>>           
>>>> AFAIAC, the post-install log is exactly where the message belongs --
>>>> that's where I'd look if I'd broken my system.
>>>>         
>>> Would it be better if your system wasn't broken?
>>>       
>> Yes, but I continue not to believe that it should be portage's job to
>> prevent me from installing things that break my system.
>>     
>
> You must be pretty unhappy with Gentoo, because portage seems
> to go to a great deal of effort to avoid breaking things (what
> with all that dependancy stuff it does).  Several times a month
> it refuses to update because of blockages alone.
>
>   

I bet with all the good work the devs do, this could be dealt with
pretty easily.  After all, they made portage so they can move
mountains.  LOL

I do think that emerging a package that will knowingly break something
is a bad idea.  I still say that if this was baselayout or some critical
package needed to boot, this would have to be dealt with quickly.  I
just don't think the devs would intentionally release a bad critical
package that is known to break something. 

Dale

:-)  :-) 

[gentoo-user] nwmouse

[gmail]

Hi!

I wonder if anyone can provide a url for me to fetch *nwmouse* The issue
is already stated on bugs:

http://bugs.gentoo.org/show_bug.cgi?id=251379

But I still need a working link to get the "cursors.tar.gz".

Thank ya!

Re: [gentoo-user] nwmouse

[Peter Humphrey]

On Saturday 03 January 2009 09:55:33 gmail wrote:
> Hi!
>
> I wonder if anyone can provide a url for me to fetch *nwmouse* The issue
> is already stated on bugs:
>
> http://bugs.gentoo.org/show_bug.cgi?id=251379
>
> But I still need a working link to get the "cursors.tar.gz".
>
> Thank ya!

In what sense is this a reply to Q's message?

-- 
Rgds
Peter