[gentoo-dev] ideas/bugs

[gentoo-dev] ideas/bugs

[Christian Loitsch]

hi,

I just finished compiling the base system on an MMX-130 Mhz with 48MB of
Ram.  (slow but doable).  Note that 48MB were nearly always enough for
the compilation!  (ok glibc would have liked more ;).

I made some observations, and would like to know, which of them are
considered as bugs.  (so that I can enter them in bugzilla)

* why is there an mta-mysql USE flag, and no mta-postgres?
* why do you call the postgresql-flag postgres?  After all its name is
  postgresql (you don't call mysql mys)
* why are the permissions of /etc/make.conf so strict (only 600 on my
  machine)  (i hope it wasn't me who changed them ;)
* as I am behind a firewall, which asks for username and password, I
  could not use wget.  The only program, which is (afaik!) capable of
  even "tunneling" through an ftp-proxy is lftp.
  I downloaded a static version (and modified the lftp-get script, so
  that it accepted a Path-option) and it worked really well.
  But it took me quite some time to realize that bootstrap doesn't use
  FETCHCOMMAND.  You could at least mention it somewhere.
  I also really think that lftp is more appropriate than wget.
* fortunately I have access to a machine outside the firewall and was
  therefore able to make the rsync there and then download it via ftp,
  would be really nice, to do this directly.
* why does the installation of grub won't work if /boot is not a
  partition on its own?  (had no cdrom and "installed" the iso-image
  temporarely on the boot-partition)
* my installed system works "worse" than the iso-image.
  the installation-manual does not explain how to set up the right
  keyboard and "make menuconfig" shows funny chars instead of the lines.
  (afaik this is only an Env. variable, but which one)
* depscan is maybe not called with the right params in the boot-scripts:
  depscan: Usage depscan ....
* why do you ask users in you manual to execute depmod, if it is in 1 of
  your boot-scripts?
* I guess the best way to avoid "broken" Gentoos because of forgetting
  the -w option with nano, would be to make an alias right at the
  beginning.  (like nanow)
* i think that you should also treat pcmcia a little bit in your
  install-manual.
* because of wrong parameters the pcmcia-boot-script failed.  But there
  was no message.  like [ FAIL ]


* I have not yet verified the next point:
  I have a pcmcia-network-card, but because of a race condition (will
  make a bug-report), I had to remove the net.eth0 script (rc-update del
  net.eth0), and call it manually.
  Surprisingly when I halt the system there is still a message:
  ??  Bringing down eth0  ??  which of course fails, because I brought
  it down already manually.


Please note that I do _not_ consider everything I mentioned as bug!


christian

ps: Gentoo really rocks!

Re: [gentoo-dev] ideas/bugs

[Martin Schlemmer]

[-- Attachment #1: Type: text/plain, Size: 4204 bytes --]

On Fri, 2002-03-01 at 20:39, Christian Loitsch wrote:
> hi,
> 
> I just finished compiling the base system on an MMX-130 Mhz with 48MB of
> Ram.  (slow but doable).  Note that 48MB were nearly always enough for
> the compilation!  (ok glibc would have liked more ;).
> 
> I made some observations, and would like to know, which of them are
> considered as bugs.  (so that I can enter them in bugzilla)
> 
> * why is there an mta-mysql USE flag, and no mta-postgres?

I'm thinking there should not be a mta-postgres one, as the
MTA should compile in either if 'mysql' or 'postgres' is in
the USE.

> * why do you call the postgresql-flag postgres?  After all its name is
>   postgresql (you don't call mysql mys)

Shortness ?  Not sure on this one.

> * why are the permissions of /etc/make.conf so strict (only 600 on my
>   machine)  (i hope it wasn't me who changed them ;)

Mine is 0644.  Should not really make a difference until
user 'emerging' is implemented.

> * as I am behind a firewall, which asks for username and password, I
>   could not use wget.  The only program, which is (afaik!) capable of
>   even "tunneling" through an ftp-proxy is lftp.
>   I downloaded a static version (and modified the lftp-get script, so
>   that it accepted a Path-option) and it worked really well.
>   But it took me quite some time to realize that bootstrap doesn't use
>   FETCHCOMMAND.  You could at least mention it somewhere.
>   I also really think that lftp is more appropriate than wget.
> * fortunately I have access to a machine outside the firewall and was
>   therefore able to make the rsync there and then download it via ftp,
>   would be really nice, to do this directly.
> * why does the installation of grub won't work if /boot is not a
>   partition on its own?  (had no cdrom and "installed" the iso-image
>   temporarely on the boot-partition)

I think it is a oversite/bug of a recent bug fix, will have
a look later on if I get the time.

> * my installed system works "worse" than the iso-image.
>   the installation-manual does not explain how to set up the right
>   keyboard and "make menuconfig" shows funny chars instead of the lines.
>   (afaik this is only an Env. variable, but which one)

You should edit /etc/rc.conf.  Check the Desktop guide, as I
am almost sertain its in there.

> * depscan is maybe not called with the right params in the boot-scripts:
>   depscan: Usage depscan ....

This probibly means you have a old/broken script in
/etc/init.d.  The next baselayout should fix this issue.

> * why do you ask users in you manual to execute depmod, if it is in 1 of
>   your boot-scripts?

Extra info/help that was included in the rc-scripts later on
to make life more painless.  Cannot hurt to remind a user,
but guess we can remove it if its really that a big a deal.

> * I guess the best way to avoid "broken" Gentoos because of forgetting
>   the -w option with nano, would be to make an alias right at the
>   beginning.  (like nanow)

Gentoo is in general for system admin's and developers
intended (last time I checked), this we try to educate, rather
than spoon feed.

> * i think that you should also treat pcmcia a little bit in your
>   install-manual.

On the TODO list somewhere.

> * because of wrong parameters the pcmcia-boot-script failed.  But there
>   was no message.  like [ FAIL ]
> 

Submit a bug report at http://bugs.gentoo.org/

> 
> * I have not yet verified the next point:
>   I have a pcmcia-network-card, but because of a race condition (will
>   make a bug-report), I had to remove the net.eth0 script (rc-update del
>   net.eth0), and call it manually.
>   Surprisingly when I halt the system there is still a message:
>   ??  Bringing down eth0  ??  which of course fails, because I brought
>   it down already manually.
> 

You sure you did a '/etc/init.d/net.eth0 stop' as well ?
What exactly is the race condition ?

> 
> Please note that I do _not_ consider everything I mentioned as bug!
> 

Sure :)



Greetings,

-- 

Martin Schlemmer
Gentoo Linux Developer, Desktop Team Developer
Cape Town, South Africa


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 232 bytes --]

[gentoo-dev] Re: ideas/bugs

[Christian Loitsch]

here some updates:

After some grepping I found the /etc/rc.conf file.
I then made some research on the net to find the best consolefonts, and
read in the french-howto that I should use the iso15-latin9 font.
Couldn't find it :(
On freshmeat.net/kbd somebody mentioned that kbd was more up to date
then console-tools and surprise surprise: kbd has this font (at least I
think it's the right one: lat9w-16).

> * depscan is maybe not called with the right params in the boot-scripts:
>   depscan: Usage depscan ....
Even though it gives this usage-message, the script seems to be executed
without any errors.

> * why do you ask users in you manual to execute depmod, if it is in 1 of
>   your boot-scripts?
it is not a big deal, but at least I did not print the manual, but read
it with less.  And when I rebootet (of course a little bit nervous ;),
it just made me even more nervous to focus on not forgetting to execute
something later on...
> * I guess the best way to avoid "broken" Gentoos because of forgetting
>   the -w option with nano, would be to make an alias right at the
>   beginning.  (like nanow)
I agree with Martin that in that it's better to educate, but let's be
honest, who uses nano?  At least all editors I use normally (ok, that's
only vim ;)  don't have this problem.

> * because of wrong parameters the pcmcia-boot-script failed.  But there
>   was no message.  like [ FAIL ]
hope nobody got this wrong.  The reason the script failed, was that I
moved the modules to another place, because I did not (and still don't)
understand why cardmgr... loaded modules I compiled into the kernel. (
as stars and not as M odules)
Nevertheless there is something wrong with the PCMCIA-package.

1) cardmgr is a daemon and because my machine is sooooo slow, it doesn't
have enough time to load all modules for my network-card.
If I add a sleep 2 inside my net.eth0 everything works fine.
I have _no_ idea how this could easily be fixed.  (unless you let
cardmgr call the right scripts, which of course is correct, but "ugly")
2) the pcmcia "-script" , even though in /etc/runlevels/default, does
not produce a link in /dev/shm/.init.d/started.  There is a directory in
/dev/shm/.init.d/use/pcmcia containing a link to net.eth0 though.
3) obviously because of (2) the pcmcia script is called twice.  Once for
the net.eth0 script and another time after (on the screen even the last
one)
4) the pcmcia-script does not "produce" a green star and no ok.
It also is closed because of sending the term-signal to everyone
and not because of calling pcmcia stop.

Ignore my last untested point! (from the original mail)
Everything is fine.  I did not imagine that the scripts were this
clever, and would insert them into /dev/shm/.init.d/started, even
when called manually.

At the moment I tend to fill out 2 bug-reports:
1) no ok for the pcmcia-script
2) race-condition

Anything else?

Christian

Re: [gentoo-dev] Re: ideas/bugs

[Martin Schlemmer]

On Mon, 2002-03-04 at 15:34, Christian Loitsch wrote:
> here some updates:
> 
> After some grepping I found the /etc/rc.conf file.
> I then made some research on the net to find the best consolefonts, and
> read in the french-howto that I should use the iso15-latin9 font.
> Couldn't find it :(
> On freshmeat.net/kbd somebody mentioned that kbd was more up to date
> then console-tools and surprise surprise: kbd has this font (at least I
> think it's the right one: lat9w-16).
> 
> > * depscan is maybe not called with the right params in the boot-scripts:
> >   depscan: Usage depscan ....
> Even though it gives this usage-message, the script seems to be executed
> without any errors.
> 
> > * why do you ask users in you manual to execute depmod, if it is in 1 of
> >   your boot-scripts?
> it is not a big deal, but at least I did not print the manual, but read
> it with less.  And when I rebootet (of course a little bit nervous ;),
> it just made me even more nervous to focus on not forgetting to execute
> something later on...
> > * I guess the best way to avoid "broken" Gentoos because of forgetting
> >   the -w option with nano, would be to make an alias right at the
> >   beginning.  (like nanow)
> I agree with Martin that in that it's better to educate, but let's be
> honest, who uses nano?  At least all editors I use normally (ok, that's
> only vim ;)  don't have this problem.
> 
> > * because of wrong parameters the pcmcia-boot-script failed.  But there
> >   was no message.  like [ FAIL ]
> hope nobody got this wrong.  The reason the script failed, was that I
> moved the modules to another place, because I did not (and still don't)
> understand why cardmgr... loaded modules I compiled into the kernel. (
> as stars and not as M odules)
> Nevertheless there is something wrong with the PCMCIA-package.
> 
> 1) cardmgr is a daemon and because my machine is sooooo slow, it doesn't
> have enough time to load all modules for my network-card.
> If I add a sleep 2 inside my net.eth0 everything works fine.
> I have _no_ idea how this could easily be fixed.  (unless you let
> cardmgr call the right scripts, which of course is correct, but "ugly")
> 2) the pcmcia "-script" , even though in /etc/runlevels/default, does
> not produce a link in /dev/shm/.init.d/started.  There is a directory in
> /dev/shm/.init.d/use/pcmcia containing a link to net.eth0 though.
> 3) obviously because of (2) the pcmcia script is called twice.  Once for
> the net.eth0 script and another time after (on the screen even the last
> one)

Strange one.  It should create the /dev/shm/.init.d/started/pcmcia link
if the script executed properly.  If it do not have the green ok, it
means 2 things: 1) no 'eend' is present in the script. 2) it fails and
should give you a red '!!'

Other posiblity is a bug in the runscript.sh script.  It however sounds
like the script fails, as there was a bug with the previous baselayout
that caused a script to start multiple times if it keeps failing. This
is fixed in the latest baselayout.

So, back to the pcmcia script.  I can check it for syntax errors, but
the how and innerworkings of pcmcia is not known to me as I do not use
a notebook.

If possible, update to the latest baselayout (1.7.2 .. maybe 1.7.2-r1 or
1.7.3 that should come out tonight), and give feedback again.  If you
can debug the problem, so the better.


> 4) the pcmcia-script does not "produce" a green star and no ok.
> It also is closed because of sending the term-signal to everyone
> and not because of calling pcmcia stop.
> 
> Ignore my last untested point! (from the original mail)
> Everything is fine.  I did not imagine that the scripts were this
> clever, and would insert them into /dev/shm/.init.d/started, even
> when called manually.
> 
> At the moment I tend to fill out 2 bug-reports:
> 1) no ok for the pcmcia-script
> 2) race-condition
> 
> Anything else?
> 
> Christian
> 
> _______________________________________________
> gentoo-dev mailing list
> gentoo-dev@gentoo.org
> http://lists.gentoo.org/mailman/listinfo/gentoo-dev
-- 
Martin Schlemmer
Gentoo Linux Developer, Desktop Team
Cape Town, South Africa

Re: [gentoo-dev] Re: ideas/bugs

[Chad M. Huneycutt]

Martin Schlemmer wrote:
> So, back to the pcmcia script.  I can check it for syntax errors, but
> the how and innerworkings of pcmcia is not known to me as I do not use
> a notebook.


Well, at one point, I had the pcmcia stuff working in portage, but I guess 
someone blew away my changes.

-- 
Chad Huneycutt                                try { Windows }
Ph.D. Student                                 catch ( Exception BSOD )
Georgia Tech College of Computing               { linux };
http://www.cc.gatech.edu/~chadh

[gentoo-dev] Re: ideas/bugs

[Christian Loitsch]

Even though I don't understand why, I fixed the pcmcia-package and
submitted a bug report.  (972)

The gentoo /etc/init.d/pcmcia script was never used, but the script
provided by the pcmcia-package.
In the pcmcia-cs...ebuild script newins was called with only one
parameter.  If somebody could explain me, why this also prevented a
sub-sequent call of newexe to succeed, please tell me.

cu
cl