[gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Walter Dnes]

  I have a main machine and a backup machine.  The main machine is 64-bit
and the backup is 32-bit, but otherwise very similar setup.  I
haven't updated the backup (32-bit machine) for a while, and decided to
do so today (115 packages emerge).  I normally...

emerge -pv --deep --update world | less

...before updating, to check for booby-traps.  Today, the output on the
backup machine blasted to screen, and did not stop until finished.
Meanwhile, the bottom of the screen shows "lines 1-4/4 (END) ".
Actually, it's 151 lines of output.  Similarly...

emerge -pv --deep --update world > x

...shows only 4 lines in file x, while the output blasted to the screen.

emerge -pv --deep --update world > x 2>&1

...does work, sending everything to file x.  That's how I know there are
151 lines of output in file x and output is going to stderr.  This
happens on only the backup machine, not the main machine.  It's not a
showstopper, and I can work around it.  Having said that, I'd like to do
a proper fix.

  The main quirk on my mcahines is that I start USE with "-*" on all my
machines.  This goes back to when the developers "in their infinite
wisdom", decided to make ipv6 a default USE flag.  It was not fun
watching Firefox and mplayer sit for 45 seconds until the IPV6 contact
attempt time out, before contacting remote sites with IPV4.

The machine is a bog-standard Dell D530 dual core. /proc/cpuinfo shows

> vendor_id       : GenuineIntel
> cpu family      : 6
> model           : 15
> model name      : Genuine Intel(R) CPU            2140  @ 1.60GHz
> stepping        : 2

  Any ideas?

-- 
Walter Dnes <waltdnes@waltdnes.org>

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Adam Carter]

[-- Attachment #1: Type: text/plain, Size: 959 bytes --]

> do so today (115 packages emerge).  I normally...
>
> emerge -pv --deep --update world | less
>
> ...before updating, to check for booby-traps.  Today, the output on the
> backup machine blasted to screen, and did not stop until finished.
> Meanwhile, the bottom of the screen shows "lines 1-4/4 (END) ".
> Actually, it's 151 lines of output.  Similarly...
>
> emerge -pv --deep --update world > x
>
> ...shows only 4 lines in file x, while the output blasted to the screen.
>
> emerge -pv --deep --update world > x 2>&1
>
> ...does work, sending everything to file x.  That's how I know there are
> 151 lines of output in file x and output is going to stderr.  This
> happens on only the backup machine, not the main machine.  It's not a
> showstopper, and I can work around it.  Having said that, I'd like to do
> a proper fix.
>

My first guess would be different versions of emerge sending output to
different places. Are the portage versions the same?

[-- Attachment #2: Type: text/html, Size: 1247 bytes --]

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Stroller]

On 17/11/2010, at 12:25am, Walter Dnes wrote:
>  I have a main machine and a backup machine.  The main machine is 64-bit
> and the backup is 32-bit, but otherwise very similar setup.  I
> haven't updated the backup (32-bit machine) for a while, 
> ... there are
> 151 lines of output in file x and output is going to stderr.  This
> happens on only the backup machine, not the main machine.

It's probably fixed in one of the updates! Why are you posting before you checked that!?!?

Use `screen` or `tmux` in place of `less` if you need to scroll back over the results of --pretend before running the emerge.  

> The main quirk on my mcahines is that I start USE with "-*" on all my
> machines.  This goes back to when the developers "in their infinite
> wisdom", decided to make ipv6 a default USE flag.

Why not just use -ipv6 as a global USE flag!?!? 

I have to admit that the Gentoo devs have in the past made decisions which have caused me to be suspicious of their sanity. But if I disagree with them over a USE flag I just add it to make.conf.

Stroller.

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Walter Dnes]

On Wed, Nov 17, 2010 at 01:41:45PM +0000, Stroller wrote

> It's probably fixed in one of the updates! Why are you posting before
> you checked that!?!?

  It started happening suddenly  And yes, I did search bugzilla,
although I admit my searches aren't always perfect.  And if "emerge
--sync"  output ends up telling me there is a portage update, I do run
it first.

> > The main quirk on my mcahines is that I start USE with "-*" on all my
> > machines.  This goes back to when the developers "in their infinite
> > wisdom", decided to make ipv6 a default USE flag.
> 
> Why not just use -ipv6 as a global USE flag!?!? 
> 
> I have to admit that the Gentoo devs have in the past made decisions
> which have caused me to be suspicious of their sanity. But if I
> disagree with them over a USE flag I just add it to make.conf.

  Because I don't want a repeat of the ipv6 fiasco where I had an almost
non-functional browser, mediaplayer (for internet files), etc, etc.  And
I also had to run "emerge --newuse --update world" and inspect the
output from "emerge -pv --depclean" and remove additional stuff, and
then revdep-rebuild to clean up the resulting extra "goodies".  If I
don't use "-*" what's the next flag that the devs will add?  And how many
of my current packages will link against it?

  On occasion, "emerge -pv --deep --update world" will complain that a
certain USE flag is required for my config.  At that time, I will decide
if I really want the package that requires the flag, and if so, whether
to enable the flag globally or one-off in /etc/package.use

  OK, so I'm a control freak.  That's one reason I left Windows.  I want
to be in charge of my machine.

-- 
Walter Dnes <waltdnes@waltdnes.org>

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Alan McKinnon]

Apparently, though unproven, at 02:20 on Thursday 18 November 2010, Walter 
Dnes did opine thusly:

>   Because I don't want a repeat of the ipv6 fiasco where I had an almost
> non-functional browser, mediaplayer (for internet files), etc, etc.  And
> I also had to run "emerge --newuse --update world" and inspect the
> output from "emerge -pv --depclean" and remove additional stuff, and
> then revdep-rebuild to clean up the resulting extra "goodies".  If I
> don't use "-*" what's the next flag that the devs will add?  And how many
> of my current packages will link against it?

emerge -avuND world will show you, in colour, USE flags that have changed. You 
can then decide what to do about them.

Your way, you have to explicitly add back in all the flags you want. You will 
not receive the benefit of seeing changed defaults (and there might be a good 
reason for the change, but now you will miss them). You also just trashed most 
of the usefulness of profiles and have to manually tracked all default USE 
changes yourself.

The usual way (not to do what you do) still lets you control as much as you 
want but with the minimum effort as opposed to the maximum effort. You will 
only need to make a decision when a decision needs to be made.

Why are you making more work for yourself? Portage is software, let it do what 
software is good at - removing drudge work from your life so you can get on 
with the important things, stuff that needs thought


-- 
alan dot mckinnon at gmail dot com

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Stroller]

On 18/11/2010, at 12:20am, Walter Dnes wrote:
> On Wed, Nov 17, 2010 at 01:41:45PM +0000, Stroller wrote
> 
>> It's probably fixed in one of the updates! Why are you posting before
>> you checked that!?!?
> 
>  It started happening suddenly  And yes, I did search bugzilla,
> although I admit my searches aren't always perfect.  And if "emerge
> --sync"  output ends up telling me there is a portage update, I do run
> it first.

You can't properly check if your bug is fixed in the updates by checking bugzilla. You need to check by *applying* the updates.

In your previous message you didn't tell us that you'd upgraded Portage to the latest before getting this error on 115 remaining updates, and if you did actually do that, it doesn't really matter, because you still need to run the other updates first.

If you phoned HP or Dell tech support and said you had a problem, they would ask if you had the latest version(s) of the software on your machine. They might well expect you to run system update to ensure all system packages are up to date, too. You shouldn't expect more support effort from the list than you would from a company you're paying for help! You might *get* more support effort from the folks here on the list, but that's because we're nice people (well, we all are except me and Alan) - make it easier for us!

I'm sorry if this comes across as rude, but I think it's daft to ask the list for help when your system isn't up to date. It might not fix the problem on this occasion, but don't sneer at me if that's the case because plenty of times it will. You've got more than one easy workaround that will overcome this problem and allow you to review the `emerge --pretend` output for the duration of this emerge. The effort of writing an email asking for help (and giving all the information) is so much more than that of just running the update, that even if the update doesn't fix the problem 95% of the time, it's still less effort cumulatively to always update the system before looking into the problem!

Incidentally: in your original message you stated that you pipe the output of `emerge -pv --update` to less. This will remove colour from  the output, cause it to be rendered in black and white, and make it harder to read. I have one machine that doesn't do terminal colours - I don't use it that much, so I'm not sure that it's worth fixing, but only the other day I was marvelling at how much easier it is to read stuff on my Gentoo boxes with with syntax highlighting sprinkled throughout my terminal. For instance I use `export MANPAGER="/usr/bin/most"`. If you're not using a GUI terminal emulator with a scrollbar, then may I respectfully suggest you install `tmux` (a replacement for GNU `screen`) and use it. It takes a little while to get familiar with it, and with its keybindings and stuff, and perhaps even to get into the habit and mindset of using it, but it really is brilliant, and it will allow you to page through output of `emerge --pretend` whilst retaining the colours that portage applies to the USE flag information.

Stroller.

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 570 bytes --]

On Thu, 18 Nov 2010 04:46:16 +0000, Stroller wrote:

> If you're not using a GUI terminal emulator with a scrollbar, then may
> I respectfully suggest you install `tmux` (a replacement for GNU
> `screen`) and use it. It takes a little while to get familiar with it,
> and with its keybindings and stuff, and perhaps even to get into the
> habit and mindset of using it, but it really is brilliant,

How does it differ from screen? Is it sufficiently better to relearn
keystrokes etc?


-- 
Neil Bothwick

If you think talk is cheap, try hiring a lawyer.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Vincent Launchbury]

On 2010/11/18 06:25AM, Neil Bothwick wrote:
> How does it differ from screen? Is it sufficiently better to relearn
> keystrokes etc?

The major difference is that you can split the screen into panes,
showing multiple ptys at the same time (similar to vim's :[v]sp, layout
wise). Besides that, I think they have mostly similar functionality.

Re: [gentoo-user] tmux vs. screen

[Stroller]

On 18/11/2010, at 11:25am, Neil Bothwick wrote:
> On Thu, 18 Nov 2010 04:46:16 +0000, Stroller wrote:
> 
>> If you're not using a GUI terminal emulator with a scrollbar, then may
>> I respectfully suggest you install `tmux` (a replacement for GNU
>> `screen`) and use it. It takes a little while to get familiar with it,
>> and with its keybindings and stuff, and perhaps even to get into the
>> habit and mindset of using it, but it really is brilliant,
> 
> How does it differ from screen? Is it sufficiently better to relearn
> keystrokes etc?

I wondered if I'd get asked this. The fundamental feature, I think, is that you can move a window from one session to another. This isn't something I've needed yet, but it seems like it could be useful if you need it.

tmux has panes within windows - I'm not sure if screen has that? tmux is supposed to be more efficient, better architected and use a fraction of the resources of screen.

tmux doesn't have an equivalent of `screen -Rd`, which tells screen to "start working in an existing session, or create a new one if that doesn't work". So I have `tmux a -d || tmux` in my Bash scrollback buffer, and run it as soon as I log on.

You absolutely *have* to change tmux's default control key. It was set to ctrl-b to avoid collisions with screen's ctrl-a when the author was first developing it, when he was still using screen himself, and it has never changed. I originally wanted to "stay with the default bindings" so that I'd be familiar with them "just in case I ever sometime had to log on to a different system without a personalised .tmux.conf". It drove me crazy very quickly - because "b" is on the wrong side of the (qwerty) keyboard you can't activate tmux controls one-handed until you change it back to ctrl-a. Obviously it's very easy to change the bindings. I will hereafter refer to ctrl-a as the default and as if it were the default, because that's what everyone immediately changes it to.

Because of tmux's architecture you can change options "on the fly" in a way that (I think) you can't with screen. If I want to change the key binding I can just type `tmux set  -g prefix C-a` (or `tmux set  -g prefix C-b`) and the change happens immediately. I think this is because the tmux command is not starting the tmux program, but instead talking to the existing tmux server (which was started when I ran just `tmux` with no parameters). Oh, you can apply changes only to particular sessions or windows - the "-g" before stands for global, but I think you could have different sessions as different colours, if you wanted to.

Likewise, with screen you can press ctrl-a then the " key to show what windows you have open. With tmux this is `ctrl-a w`, but you can also type `tmux list-windows` (or `tmux list-w` for short) at your bash prompt and get the same thing on stdout. You could grep that if you wanted to and I think you could use that output to make functions or bash scripts that act upon it. You can do the `ctrl-a :` thing first and enter the tmux command-line if you want to - in that case it would be `ctrl-a :list-windows` without the `tmux` in front - but you don't have to, and I find it clearer and more useful to have the output on stdout.

I believe the scrollback buffer on tmux is better, and that you can copy and paste from it in ways that you can't with screen. The default key binding to enter scrollback is `ctrl-a [`, instead of `ctrl-a esc` - I'm not entirely sure how I feel about that yet. Navigating in scrollback you can choose between vi and emacs keybindings (for page up, page down &c).

I think what really won my heart, and did so quite quickly, is that tmux has a status bar configured by default. I'm pretty sure you can do that with screen, too, but I've never bothered, because it seemed too much effort to learn and it just seemed flashy and pointless. I realised how mistaken I was within a couple of hours of using tmux. It has absolutely changed the way I use terminal multiplexers, and so I spent several hours the next day configuring mine and getting the colours and stuff perfect. 

The status bar makes navigating between windows much easier because you can see at a glance which window is which, and what's going on in them. If you don't name a window then tmux will try to show something sensible for the window title, so they show "emerge" or "vi" or "sudo" or "bash" without you having to do anything. And this will change, when the command finishes and you run something else, so basically naming windows becomes redundant and less useful. tmux will also do slightly intelligent searching, so you type `ctrl-a f` then `eme` it will suggest the window in which emerge is running.

But mostly I don't need to do tricks, because I have the status bar there showing me what's happening in which window, so I tend just to switch to windows numerically. I think having the status bar actually kinda gives you a better "spatial orientation" or a mental representation or something of the virtual windows, and it becomes more intuitive to know which one is "next" to another and where they are in relation to each other.

tmux has `ctrl-a l` to switch between two windows, a bit like if you used `cd -` a lot in the shell; you can just `ctrl-a l` repeatedly to switch between two windows (say the one you're working in and the one you're viewing results in), then you can `ctrl-a 7` to go to a third window. tmux has `ctrl-a p` and `ctrl-a n`, of course, but in screen that tended to be my *major* way of navigating, and I think that's because I didn't know where I was; now it's less so, and I feel like I'm using tmux better than I used screen. Also, tmux has a shortcut bound by default for renumbering windows, `ctrl-a .`. So say I'm working in mostly windows 3, 4 & 5, and I start a process in one of them which is going to take a while - I just `ctrl-a .9` and the window is "thrown out of the way" and becomes window 9, which resides on the right-hand side of the status bar. I can now `ctrl-a p` and `ctrl-a n` between the two remaining windows, or I can `ctrl-a c` to create a new one, which will probably fall in the gap I just left (but if not I can renumber it easily).

I've rambled on a bit here, and I guess to some people who have never used screen in the first place it'll all sound a bit WTF (if they're not familiar with the terminal-emulator virtual windows idea that screen originally introduced) and quite possibly if you're a screen power user you might be able to point out inaccuracies in what I've said, or the improvements may sound minor, or they're things that you've already configured in your own .screenrc

The TL;DR, though, is that I think it's very sensible of tmux to have a status bar by default and quite an oversight of screen not to, and that tmux is generally better than screen in ways that are quite small and subtle when taken separately, but which all together make for a nicer experience. To me it feels like a *much* nicer experience and I get the feeling that I'm not alone - there's a thread about this on the Arch forums <https://bbs.archlinux.org/viewtopic.php?id=84157>. Also: <http://news.ycombinator.com/item?id=1098564>.

Stroller.

Re: [gentoo-user] tmux vs. screen

[Paul Hartman]

On Thu, Nov 18, 2010 at 8:27 AM, Stroller
<stroller@stellar.eclipse.co.uk> wrote:
> I think what really won my heart, and did so quite quickly, is that tmux has a
>status bar configured by default. I'm pretty sure you can do that with screen,
>too, but I've never bothered, because it seemed too much effort to learn and it
>just seemed flashy and pointless. I realised how mistaken I was within a couple
>of hours of using tmux. It has absolutely changed the way I use terminal
>multiplexers, and so I spent several hours the next day configuring mine and
>getting the colours and stuff perfect.

I have not used tmux but I agree completely, I hate to use screen
without the status bar. I'm using one I copied from here or the forums
or the gentoo wiki or someplace out there in WWW land. (Thanks to the
person who made it, whoever you are)

Add this to your .screenrc:
caption always "%{= kw}%-w%{= BW}%n %t%{-}%+w %-= @%H - %LD %d %LM - %c"

Re: [gentoo-user] tmux vs. screen

[Florian CROUZAT]

On 18 nov. 2010, at 20:52, Paul Hartman wrote:

> On Thu, Nov 18, 2010 at 8:27 AM, Stroller
> <stroller@stellar.eclipse.co.uk> wrote:
>> I think what really won my heart, and did so quite quickly, is that tmux has a
>> status bar configured by default. I'm pretty sure you can do that with screen,
>> too, but I've never bothered, because it seemed too much effort to learn and it
>> just seemed flashy and pointless. I realised how mistaken I was within a couple
>> of hours of using tmux. It has absolutely changed the way I use terminal
>> multiplexers, and so I spent several hours the next day configuring mine and
>> getting the colours and stuff perfect.
> 
> I have not used tmux but I agree completely, I hate to use screen
> without the status bar. I'm using one I copied from here or the forums
> or the gentoo wiki or someplace out there in WWW land. (Thanks to the
> person who made it, whoever you are)
> 
> Add this to your .screenrc:
> caption always "%{= kw}%-w%{= BW}%n %t%{-}%+w %-= @%H - %LD %d %LM - %c"

Also, amongst other things, >=tmux-1.3 has mouse support.
You can scroll using your mouse in copy mode and use your mouse to select one of the splitted panes of your active window.
You can also break/join panes in and out the active window and it has awesome predefined layouts.
I'll add that tmux has a readable and even understandable man page and the dev(s) is really reactive on the ml/irc.

ps : in case anyone needs a customized tmux.conf, here's mine http://sprunge.us/WHNU

-----
Florian.
/ For security reasons, all text in this mail 
  is double-rot13 encrypted. /

Re: [gentoo-user] tmux vs. screen

[Alan McKinnon]

Apparently, though unproven, at 00:25 on Friday 19 November 2010, Florian 
CROUZAT did opine thusly:

> On 18 nov. 2010, at 20:52, Paul Hartman wrote:
> > On Thu, Nov 18, 2010 at 8:27 AM, Stroller
> > 
> > <stroller@stellar.eclipse.co.uk> wrote:
> >> I think what really won my heart, and did so quite quickly, is that tmux
> >> has a status bar configured by default. I'm pretty sure you can do that
> >> with screen, too, but I've never bothered, because it seemed too much
> >> effort to learn and it just seemed flashy and pointless. I realised how
> >> mistaken I was within a couple of hours of using tmux. It has
> >> absolutely changed the way I use terminal multiplexers, and so I spent
> >> several hours the next day configuring mine and getting the colours and
> >> stuff perfect.
> > 
> > I have not used tmux but I agree completely, I hate to use screen
> > without the status bar. I'm using one I copied from here or the forums
> > or the gentoo wiki or someplace out there in WWW land. (Thanks to the
> > person who made it, whoever you are)
> > 
> > Add this to your .screenrc:
> > caption always "%{= kw}%-w%{= BW}%n %t%{-}%+w %-= @%H - %LD %d %LM - %c"
> 
> Also, amongst other things, >=tmux-1.3 has mouse support.
> You can scroll using your mouse in copy mode and use your mouse to select
> one of the splitted panes of your active window. You can also break/join
> panes in and out the active window and it has awesome predefined layouts.
> I'll add that tmux has a readable and even understandable man page
                           ^^^^^^^^          ^^^^^^^^^^^^^^ ^^^ ^^^^

I'm sold. 4 words, that's all it took. 

-- 
alan dot mckinnon at gmail dot com

Re: [gentoo-user] tmux vs. screen

[Dale]

Alan McKinnon wrote:
> Apparently, though unproven, at 00:25 on Friday 19 November 2010, Florian
> CROUZAT did opine thusly:
>
>    
>> On 18 nov. 2010, at 20:52, Paul Hartman wrote:
>>      
>>> On Thu, Nov 18, 2010 at 8:27 AM, Stroller
>>>
>>> <stroller@stellar.eclipse.co.uk>  wrote:
>>>        
>>>> I think what really won my heart, and did so quite quickly, is that tmux
>>>> has a status bar configured by default. I'm pretty sure you can do that
>>>> with screen, too, but I've never bothered, because it seemed too much
>>>> effort to learn and it just seemed flashy and pointless. I realised how
>>>> mistaken I was within a couple of hours of using tmux. It has
>>>> absolutely changed the way I use terminal multiplexers, and so I spent
>>>> several hours the next day configuring mine and getting the colours and
>>>> stuff perfect.
>>>>          
>>> I have not used tmux but I agree completely, I hate to use screen
>>> without the status bar. I'm using one I copied from here or the forums
>>> or the gentoo wiki or someplace out there in WWW land. (Thanks to the
>>> person who made it, whoever you are)
>>>
>>> Add this to your .screenrc:
>>> caption always "%{= kw}%-w%{= BW}%n %t%{-}%+w %-= @%H - %LD %d %LM - %c"
>>>        
>> Also, amongst other things,>=tmux-1.3 has mouse support.
>> You can scroll using your mouse in copy mode and use your mouse to select
>> one of the splitted panes of your active window. You can also break/join
>> panes in and out the active window and it has awesome predefined layouts.
>> I'll add that tmux has a readable and even understandable man page
>>      
>                             ^^^^^^^^          ^^^^^^^^^^^^^^ ^^^ ^^^^
>
> I'm sold. 4 words, that's all it took.
>
>    

That makes me want to try it too.  Most man pages are like Greek to me.  
I use screen pretty often, especially when OOo is compiling.  Nothing 
worse than starting that thing in a Konsole and realizing you need to 
log out about 6 hours in.  o-O

Dale

:-)  :-)

Re: [gentoo-user] tmux vs. screen

[Jesús J. Guerrero Botella]

These days tmux seems somewhat more actively developed, this might be
a subjective appreciation, though. But at least the tmux mailing list
seems more active. I am subscribed to both of them.

The feature that made me switch from screen to tmux back in the days
was vertical splitting, which just worked and worked well (not at
turtle pace, not blowing my cpu like in screen). It happens that
nowadays I no longer use that feature, but I kept tmux because I find
it simpler to handle, because it's lighter in my RAM, and because a
few problems with screen (besides splitting) that I no longer
remember. I switched long ago.

I have no idea if my concerns about screen still hold true. Not do I
care, since for now I have no need to change back.

-- 
Jesús Guerrero Botella

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 885 bytes --]

On Wed, 17 Nov 2010 19:20:25 -0500, Walter Dnes wrote:

>   Because I don't want a repeat of the ipv6 fiasco where I had an almost
> non-functional browser, mediaplayer (for internet files), etc, etc.

Hardly non-functional - desperately slow maybe but far from
non-functional. Blaming the devs for your broken modem/router is rather
unfair. If you'd known it was unable to handle IPv6 correctly, why didn't
you set the flag accordingly? If you didn't know, HTH were the devs
supposed to know?

Bear in mind the current thread discussing Gentoo following upstream
closely and setting suitable defaults. Upstream include IPv6 support,
Gentoo provides an option to turn this off, you missed it. That's a very
poor performance for a control freak (speaking from a personal
perspective).


-- 
Neil Bothwick

A man wrapped up in himself makes a very small package.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Walter Dnes]

On Thu, Nov 18, 2010 at 11:21:33AM +0000, Neil Bothwick wrote

> Blaming the devs for your broken modem/router is rather unfair. If
> you'd known it was unable to handle IPv6 correctly, why didn't you
> set the flag accordingly?

  My ISP didn't support ipv6 at that time.  They're now running a beta
for native ipv6 (no tunneling) but I don't have the time to play with
bleeding edge stuff.  Regardless of the fact that my router/modem does
or does not support ipv6, if I don't have ipv6 service from my ISP (or a
tunnel broker) ipv6 is pointless.

> If you didn't know, HTH were the devs supposed to know?

  The devs *CHANGED AN EXISTING DEFAULT FLAG* from -ipv6 to ipv6.  What
percentage of the user base was running ipv6 a couple of years ago?  Why
couldn't they have left the default at -ipv6?  Ever heard of "the
principle of least surprise" aka "the principle of least astonishment"?
See http://en.wikipedia.org/wiki/Principle_of_least_astonishment
Unnecessarily changing defaults violates that principle in the worst way.

  There's an old saying...
* good judgement is the result of experience
* experience is the result of bad judgement

  As a result of my "experience" with the ipv6 flag, I no longer
robo-update.  Note that in the first post of this thread, I said...

>   I normally...
> 
> emerge -pv --deep --update world | less
> 
> ...before updating, to check for booby-traps.

  So you see, I did learn from my "experience".  I do check for stuff
like this now.  As an additional safety measure, I also begin the USE
variable with "-*".

-- 
Walter Dnes <waltdnes@waltdnes.org>

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Dale]

Walter Dnes wrote:
>    So you see, I did learn from my "experience".  I do check for stuff
> like this now.  As an additional safety measure, I also begin the USE
> variable with "-*".
>
>    

I'm starting to like the way Walter thinks.  LOL

Dale

:-)  :-)

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1280 bytes --]

On Fri, 19 Nov 2010 01:34:15 -0500, Walter Dnes wrote:

> > Blaming the devs for your broken modem/router is rather unfair. If
> > you'd known it was unable to handle IPv6 correctly, why didn't you
> > set the flag accordingly?  
> 
>   My ISP didn't support ipv6 at that time.  They're now running a beta
> for native ipv6 (no tunneling) but I don't have the time to play with
> bleeding edge stuff.  Regardless of the fact that my router/modem does
> or does not support ipv6, if I don't have ipv6 service from my ISP (or a
> tunnel broker) ipv6 is pointless.

But harmless. The severe delays you noticed were the result of a broken
modem/router failing to recognise that IPv6 was not available and trying
to use it anyway. The usual fix for such a problem is a firmware update.

I don't use -*, I did see the ipv6 flag pop up in emerge -uaD world, I
disabled it. I supposed it comes down to how much you trust yourself and
the devs.

Incidentally, you are better off not using -v in this situation because
then only changed flags are shown. I still have to get out of the habit
of using it, gained years ago when portage didn't shown any use details
when you didn't use -v.


-- 
Neil Bothwick

Definition of Trust: Two cannibals having oral sex.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Walter Dnes]

On Fri, Nov 19, 2010 at 09:25:18AM +0000, Neil Bothwick wrote

> But harmless. The severe delays you noticed were the result of a
> broken modem/router failing to recognise that IPv6 was not available
> and trying to use it anyway. The usual fix for such a problem is a
> firmware update.

  It's more complex than that.  How is the IPV6-enabled browser or media
player supposed to know that my modem doesn't support IPV6 and neither
does my ISP and neither do umpteen hops between me and the site I'm
trying to connect to?  See http://www.ipjforum.org/?p=378

> The technology in web browsers and operating systems involves doing
> Domain Name System (DNS) queries for AAAA and A resource records and
> then attempting to connect to the resulting IPv6 and IPv4 addresses
> sequentially. If the IPv6 path is broken (or slow), this connection
> can take a long time before it falls back to trying IPv4. This process
> is especially painful on typical websites that retrieve objects
> from different hosts-each failure incurs a delay. The combination of
> operating system and web browser results in delays from 20 seconds to
> several minutes if the IPv6 path is broken[2]. The typical message
> flow of a TCP client is shown in Figure 1. Clearly, this delay is
> unacceptable to users. Users avoid this delay by disabling IPv6[3]
> or avoiding IPv6-enabled websites.

  The decision to enable IPV6 by default was a mistake.  The only
beneficial side effect was that it taught me not to do robo-updates
any more <G>.

-- 
Walter Dnes <waltdnes@waltdnes.org>

Re: [gentoo-user] One machine sends "emerge" text output to stderr, not stdout

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 994 bytes --]

On Sat, 20 Nov 2010 00:32:28 -0500, Walter Dnes wrote:

> > But harmless. The severe delays you noticed were the result of a
> > broken modem/router failing to recognise that IPv6 was not available
> > and trying to use it anyway. The usual fix for such a problem is a
> > firmware update.  
> 
>   It's more complex than that.  How is the IPV6-enabled browser or media

It's not the browser, its the resolver, the delays are caused by
excessively long DNS lookups.

> player supposed to know that my modem doesn't support IPV6

The modem tells it and the resolver falls back to IPv4. The problem
occurs when the modem doesn't know that it doesn't support IPv6 (or even
what IPv6 is, which is why a firmware update usually fixes it) so the
resolver doesn't try IPv4 until the IPv6 request times out.

Using 21st Century software with 20th Century hardware is asking for such
problems :(


-- 
Neil Bothwick

I am McCoy of Bo...Damnit! I'm a doctor, not a collective!

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]