opening new tab does not keep working directory

Forwarded from Debian and confirmed on Ubuntu 12.04

Yes, confirmed. It happens with bash, but not with zsh.

I'll take back my words. It's not confirmed. I've been testing with bash and zsh in 3.0.3. The working directory is kept when a new tab is opened. There is only one case where it doesn't work, and it's when a subshell is executed, the current working directory is changed and a new tab is opened. And that is a expected behaviour.

Andrew, you wrote that it's confirmed but i think it's the subshell behaviour i mentioned above. Please correct me if i'm wrong.

Hi David,

On Mi, 11 Apr 2012, David Gómez wrote:
> I'll take back my words. It's not confirmed. I've been testing with bash
> and zsh in 3.0.3. The working directory is kept when a new tab is
> opened. There is only one case where it doesn't work, and it's when a
> subshell is executed, the current working directory is changed and a
> new tab is opened. And that is a expected behaviour.

What do you mean with "a subshell is executed"?
Or in other words, how to fix bash so that a new tab keeps the CWD?

Best wishes

Norbert
------------------------------------------------------------------------
Norbert Preining preining@{jaist.ac.jp, logic.at, debian.org}
JAIST, Japan TeX Live & Debian Developer
DSA: 0x09C5B094 fp: 14DF 2E6C 0307 BE6D AD76 A9C0 D2BF 4AA3 09C5 B094
------------------------------------------------------------------------
VENTNOR (n.) One who, having been visited as a child by a mysterious
gypsy lady, is gifted with the strange power of being able to operate
the air-nozzles above aeroplane seats.
   --- Douglas Adams, The Meaning of Liff

Hi David,

once more me ...

On Do, 12 Apr 2012, Norbert Preining wrote:
> On Mi, 11 Apr 2012, David Gómez wrote:
> > I'll take back my words. It's not confirmed. I've been testing with bash
> > and zsh in 3.0.3. The working directory is kept when a new tab is
> > opened. There is only one case where it doesn't work, and it's when a
> > subshell is executed, the current working directory is changed and a
> > new tab is opened. And that is a expected behaviour.
>
> What do you mean with "a subshell is executed"?
> Or in other words, how to fix bash so that a new tab keeps the CWD?

I have now set "set -x" in /etc/bash.bashrc, /etc/profile and I don't
see anything specific in a new tab that draws my attention. The process
is running
 bash
without any arguments (so no login shell or so).

The output after setting the -x is:
+ '[' -z '\s-\v\$ ' ']'
+ shopt -s checkwinsize
+ '[' -z '' ']'
+ '[' -r /etc/debian_chroot ']'
+ PS1='${debian_chroot:+($debian_chroot)}\u@\h:\w\$ '
+ '[' -x /usr/lib/command-not-found -o -x /usr/share/command-not-found/command-not-found ']'
+ '[' -z '${debian_chroot:+($debian_chroot)}\u@\h:\w\$ ' ']'
+ export HISTCONTROL=ignoreboth,ignoredups
+ HISTCONTROL=ignoreboth,ignoredups
+ export HISTCONTROL=ignoreboth
+ HISTCONTROL=ignoreboth
+ shopt -s histappend
+ shopt -s checkwinsize
+ '[' -z '' ']'
+ '[' -r /etc/debian_chroot ']'
+ case "$TERM" in
+ '[' -n '' ']'
+ '[' '' = yes ']'
+ PS1='${debian_chroot:+($debian_chroot)}\u@\h:\w\$ '
+ unset color_prompt force_color_prompt
+ case "$TERM" in
+ PS1='\[\e]0;\u@\h: \w\a\]${debian_chroot:+($debian_chroot)}\u@\h:\w\$ '
+ '[' -x /usr/bin/dircolors ']'
++ dircolors -b
+ eval 'LS_COLORS='\''rs=0:di=0...............................
++ export LS_COLORS
+ alias 'ls=ls --color=auto'
+ . /etc/profile
++ set -x
+++ id -u
++ '[' 1000 -eq 0 ']'
++ PATH=/usr/local/bin:/usr/bin:/bin:/usr/games
++ '[' '\[\e]0;\u@\h: \w\a\]${debian_chroot:+($debian_chroot)}\u@\h:\w\$ ' ']'
++ '[' /bin/bash ']'
++ PS1='\u@\h:\w\$ '
++ export PATH
++ umask 022
++ '[' -r /etc/profile.local ']'
++ . /etc/profile.local
+++ export PATH
++++ lesspipe
+++ eval export 'LESSOPEN="|' /usr/bin/lesspipe '%s";' export 'LESSCLOSE="/usr/bin/lesspipe' %s '%s";'
++++ export 'LESSOPEN=| /usr/bin/lesspipe %s'
++++ LESSOPEN='| /usr/bin/lesspipe %s'
++++ export 'LESSCLOSE=/usr/bin/lesspipe %s %s'
++++ LESSCLOSE='/usr/bin/lesspipe %s %s'
+++ unset PROMPT_COMMAND
+++ PS1='\u@\h:\n[\w] '
+++ PS2='\w_ '
+++ IGNOREEOF=0
+++ alias 'ls=ls -F'
+++ alias 'la=ls -a'
+++ alias 'll=ls -l'
+++ alias 'l=ls -la'
+++ alias 'j=jobs -l'
+++ alias pu=pushd
+++ alias po=popd
+++ alias bye=exit
+++ alias more=less
+++ alias m=/bin/more
+++ alias 'unix2dos=recode latin1...latin1/cl'
+++ alias 'dos2unix=recode latin1/cl...latin1'
+++ export 'LESS=-q -q -q -q '
+++ LESS='-q -q -q -q '
+++ umask 022
+ export <email address hidden>
+ <email address hidden>
+ export 'DEBFULLNAME=Norbert Preining'
+ DEBFULLNAME='Norbert Preining'
+ export TEX_CATALOGUE=/src/TeX/texcatalogue-svn/entries
+ TEX_CATALOGUE=/src/TeX/texcatalogue-svn/entries
+ export CTAN=/src/TeX/CTAN
+ CTAN=/src/TeX/CTAN
+ export PATH=/usr/local/bin:/usr/bin:/bin:/usr/games:/src/TeX/tex...

By subshell i mean running a bash shell as a child of another bash shell. If you open a new tab, the CWD won't be kept.

The terminal i'm testing, doesn't matter what shell, it keeps the CWD in the new tabs.

Open a couple of tabs in a sakura terminal, check that the CWD is not kept and send the output of 'ps -axf' (only the part with the sakura proccess and its childs)

On Do, 12 Apr 2012, David Gómez wrote:
> By subshell i mean running a bash shell as a child of another bash
> shell. If you open a new tab, the CWD won't be kept.

Hmm, but how can you do it in a different way?

> Open a couple of tabs in a sakura terminal, check that the CWD is not
> kept and send the output of 'ps -axf' (only the part with the sakura
> proccess and its childs)

23933 ? Sl 0:00 sakura
23938 ? S 0:00 \_ gnome-pty-helper
23939 pts/0 Ss+ 0:00 \_ /bin/bash
23951 pts/2 Ss+ 0:00 \_ /bin/bash
23962 pts/3 Ss+ 0:00 \_ /bin/bash
23973 pts/4 Ss 0:00 \_ /bin/bash
23985 pts/4 R+ 0:00 \_ ps -axf

So, we have to change sakura to start a *new* shell instead of
sub-processing? Do I understand that right?

Best wishes

Norbert
------------------------------------------------------------------------
Norbert Preining preining@{jaist.ac.jp, logic.at, debian.org}
JAIST, Japan TeX Live & Debian Developer
DSA: 0x09C5B094 fp: 14DF 2E6C 0307 BE6D AD76 A9C0 D2BF 4AA3 09C5 B094
------------------------------------------------------------------------
MAPLEDURHAM (n.)
A hideous piece of chipboard veneer furniture bought in a suburban
high street furniture store and designed to hold exactly a year's
supply of Sunday colour supplements.
   --- Douglas Adams, The Meaning of Liff

No. I mean, if you open a sakura terminal, and you type "bash" (that is, a subshell) in any given tab, and under the newly created shell you open new tabs, CWD won't be kept. That's expected behaviour.

But, looking at your 'ps' output, you don't have any subshells running. I don't know why the CWD is not being kept in your system. In my tests CWD is changed correctly.

Just do one more test to see if we can find the cause. Launch a sakura terminal with just one tab, and change to a directory different from your $HOME. Sent me the output of 'ls -l /proc/pid' where pid will be the process id of the bash shell running as a child of the sakura proccess.

On Sa, 14 Apr 2012, David Gómez wrote:
> No. I mean, if you open a sakura terminal, and you type "bash" (that is,
> a subshell) in any given tab, and under the newly created shell you open
> new tabs, CWD won't be kept. That's expected behaviour.

Of course, sorry that I didn't grasp that. Yes, I agree.

> But, looking at your 'ps' output, you don't have any subshells running.
> I don't know why the CWD is not being kept in your system. In my tests
> CWD is changed correctly.

Hmmm, strange, ok ...

> Just do one more test to see if we can find the cause. Launch a sakura
> terminal with just one tab, and change to a directory different from
> your $HOME. Sent me the output of 'ls -l /proc/pid' where pid will be
> the process id of the bash shell running as a child of the sakura
> proccess.

Iam now in /tmp and typed ls -l /proc/pid and received a list of files:
[/tmp] ls -l /proc/15374
total 0
-rw-r--r-- 1 norbert norbert 0 Apr 15 21:14 autogroup
-r-------- 1 norbert norbert 0 Apr 15 21:14 auxv
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 cgroup
--w------- 1 norbert norbert 0 Apr 15 21:14 clear_refs
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 cmdline
-rw-r--r-- 1 norbert norbert 0 Apr 15 21:14 comm
-rw-r--r-- 1 norbert norbert 0 Apr 15 21:14 coredump_filter
lrwxrwxrwx 1 norbert norbert 0 Apr 15 21:14 cwd -> /tmp/
-r-------- 1 norbert norbert 0 Apr 15 21:14 environ
lrwxrwxrwx 1 norbert norbert 0 Apr 15 21:14 exe -> /bin/bash*
dr-x------ 2 norbert norbert 0 Apr 15 21:13 fd/
dr-x------ 2 norbert norbert 0 Apr 15 21:14 fdinfo/
-r-------- 1 norbert norbert 0 Apr 15 21:14 io
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 limits
-rw-r--r-- 1 norbert norbert 0 Apr 15 21:14 loginuid
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 maps
-rw------- 1 norbert norbert 0 Apr 15 21:14 mem
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 mountinfo
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 mounts
-r-------- 1 norbert norbert 0 Apr 15 21:14 mountstats
dr-xr-xr-x 4 norbert norbert 0 Apr 15 21:14 net/
dr-x--x--x 2 norbert norbert 0 Apr 15 21:14 ns/
-rw-r--r-- 1 norbert norbert 0 Apr 15 21:14 oom_adj
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 oom_score
-rw-r--r-- 1 norbert norbert 0 Apr 15 21:14 oom_score_adj
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 pagemap
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 personality
lrwxrwxrwx 1 norbert norbert 0 Apr 15 21:14 root -> //
-rw-r--r-- 1 norbert norbert 0 Apr 15 21:14 sched
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 sessionid
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 smaps
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 stack
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 stat
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 statm
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 status
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 syscall
dr-xr-xr-x 3 norbert norbert 0 Apr 15 21:14 task/
-r--r--r-- 1 norbert norbert 0 Apr 15 21:14 wchan
[/tmp]

Does this help in any way?

Best wishes

Norbert
------------------------------------------------------------------------
Norbert Preining preining@{jaist.ac.jp, logic.at, debian.org}
JAIST, Japan TeX Live & Debian Developer
DSA: 0x09C5B094 fp: 14DF 2E6C 0307 B...

I can confirm this happens under Debian with zsh as shell as well. If sakura has been started with a custom CWD the new tabs all use this custom CWD:

~ $ cd /tmp
/tmp $ sakura &
[1] PID
/tmp $

[tab 1]
/tmp $ cd /var
/var $ <STRG+SHIFT+T>

[tab 2]
/tmp $

It does not happen with gnome-terminal and evilvte, so it's probably not a bug in libvte.

On Mi, 25 Apr 2012, Sebastian Reichel wrote:
> I can confirm this happens under Debian with zsh as shell as well. If

Thanks, good to know. I have no idea what the reason is, but
it is a pain.

Best wishes

Norbert
------------------------------------------------------------------------
Norbert Preining preining@{jaist.ac.jp, logic.at, debian.org}
JAIST, Japan TeX Live & Debian Developer
DSA: 0x09C5B094 fp: 14DF 2E6C 0307 BE6D AD76 A9C0 D2BF 4AA3 09C5 B094
------------------------------------------------------------------------
AINDERBY STEEPLE (n.)
One who asks you a question with the apparent motive of wanting to
hear your answer, but who cuts short your opening sentence by leaning
forward and saying 'and I'll tell you why I ask...' and then talking
solidly for the next hour.
   --- Douglas Adams, The Meaning of Liff

Hi,

I checked the code, since this is really annoying. This error is *not* upstream's fault. The Debian patch, which fixes the HURD build [0] is broken. I suggest to fix the patch and merge it upstream afterwards. There are multiple problems with the patch:

1. It does lstat(file, &sb) before file has been filled with "/proc/<pid>/cwd", so it does lstat on NULL
2. It should do stat() instead of lstat(), since we are interested in the size of the path and not the size of the directory, which is 0
3. file must be free'd if lstat() fails
4. file must be free'd if size check fails

So the patch introduces 2 errors breaking the code for sure and two potential memory leaks.

Andrew Starr-Bochicchio, you should review patches before applying them.

David Gómez, I guess the updated patch should go upstream. Using the actual path size is a cleaner approach, than using MAX_PATH and is POSIX compatible.

[0] http://patch-tracker.debian.org/patch/series/view/sakura/2.4.2-4/fix_FTBFS4Hurd.patch

-- Sebastian

I see, that's why i wasn't able to reproduce this bug...

Sebastian, thanks for looking into this.

Honestly, i don't give a shit about hurd. But removing PATH_MAX is a good thing. Give me a couple of days to write a patch to use dynamic memory. I'll try to do it ASAP so Andrew can release a new package with this problem fixed.

I have patch ready.

-- Sebastian

Great! You're fast! :D I'll add it right now.

Hi Andrew, hi Sebastian,

I have updated the debian packaging repository of sakura with this
patch (thanks Sebastian) and prepared 3.0.3-3 packages.

Andrew, since I am also in the Uploaders field I can easily upload
this version, and push also a tag to the debian sakura.git.
Please let me know *soon* what your plans are, otherwise I will upload
rather sooner than later, maybe already tomorrow.

On So, 29 Apr 2012, Sebastian Reichel wrote:
> 1. It does lstat(file, &sb) before file has been filled with "/proc/<pid>/cwd", so it does lstat on NULL
> 2. It should do stat() instead of lstat(), since we are interested in the size of the path and not the size of the directory, which is 0
> 3. file must be free'd if lstat() fails
> 4. file must be free'd if size check fails
>
> So the patch introduces 2 errors breaking the code for sure and two
> potential memory leaks.
>
> Andrew Starr-Bochicchio, you should review patches before applying them.
>
> David Gómez, I guess the updated patch should go upstream. Using the
> actual path size is a cleaner approach, than using MAX_PATH and is POSIX
> compatible.
>
> [0] http://patch-
> tracker.debian.org/patch/series/view/sakura/2.4.2-4/fix_FTBFS4Hurd.patch
>
> -- Sebastian

Best wishes

Norbert
------------------------------------------------------------------------
Norbert Preining preining@{jaist.ac.jp, logic.at, debian.org}
JAIST, Japan TeX Live & Debian Developer
DSA: 0x09C5B094 fp: 14DF 2E6C 0307 BE6D AD76 A9C0 D2BF 4AA3 09C5 B094
------------------------------------------------------------------------
I teleported home one night
With Ron and Sid and Meg.
Ron stole Meggie's heart away
And I got Sidney's leg.
                 --- A poem about matter transference beams.
                 --- Douglas Adams, The Hitchhikers Guide to the Galaxy

Norbert,

I released sakura-3.0.4 some hours ago. I think it'd be better if you package this version which already has the patch included. Thanks.

On Di, 01 Mai 2012, David Gómez wrote:
> I released sakura-3.0.4 some hours ago. I think it'd be better if you
> package this version which already has the patch included. Thanks.

Thanks,

Andrew, orig.tar is include in pristine-tar branch, upstream branch
updated, master (debian) branch updated for 3.0.4, patch unfuzzified,
other patches disabled. package build, lintian clean.

Same as before still applies, please comment on your plans.

Best wishes

Norbert
------------------------------------------------------------------------
Norbert Preining preining@{jaist.ac.jp, logic.at, debian.org}
JAIST, Japan TeX Live & Debian Developer
DSA: 0x09C5B094 fp: 14DF 2E6C 0307 BE6D AD76 A9C0 D2BF 4AA3 09C5 B094
------------------------------------------------------------------------
GLAZELEY (adj.)
The state of a barrister's flat greasy hair after wearing a wig all
day.
   --- Douglas Adams, The Meaning of Liff

Thanks for figuring it out!

Sorry for the silence. I've been away from a computer (or even cell
reception) for the past few days.

Norbert, feel free to upload. If you don't get to it, I'll be back in front
of a computer tomorrow

Thanks,

Andrew
On Apr 30, 2012 8:57 PM, "Norbert Preining" <email address hidden> wrote:

> On Di, 01 Mai 2012, David Gómez wrote:
> > I released sakura-3.0.4 some hours ago. I think it'd be better if you
> > package this version which already has the patch included. Thanks.
>
> Thanks,
>
> Andrew, orig.tar is include in pristine-tar branch, upstream branch
> updated, master (debian) branch updated for 3.0.4, patch unfuzzified,
> other patches disabled. package build, lintian clean.
>
> Same as before still applies, please comment on your plans.
>
> Best wishes
>
> Norbert
> ------------------------------------------------------------------------
> Norbert Preining preining@{jaist.ac.jp, logic.at, debian.org}
> JAIST, Japan TeX Live & Debian Developer
> DSA: 0x09C5B094 fp: 14DF 2E6C 0307 BE6D AD76 A9C0 D2BF 4AA3 09C5 B094
> ------------------------------------------------------------------------
> GLAZELEY (adj.)
> The state of a barrister's flat greasy hair after wearing a wig all
> day.
> --- Douglas Adams, The Meaning of Liff
>

Hi Andrew,

good to hear from you.

On Di, 01 Mai 2012, Andrew Starr-Bochicchio wrote:
> Sorry for the silence. I've been away from a computer (or even cell
> reception) for the past few days.

No problem at all.

> Norbert, feel free to upload. If you don't get to it, I'll be back in front
> of a computer tomorrow

I uploaded 3.0.4-1, and made a signed tag debian/3.0.4-1, and pushed
all to the repository.

Thanks a lot

Norbert

------------------------------------------------------------------------
Norbert Preining preining@{jaist.ac.jp, logic.at, debian.org}
JAIST, Japan TeX Live & Debian Developer
DSA: 0x09C5B094 fp: 14DF 2E6C 0307 BE6D AD76 A9C0 D2BF 4AA3 09C5 B094
------------------------------------------------------------------------
WRABNESS (n.)
The feeling after having tried to dry oneself with a damp towel.
   --- Douglas Adams, The Meaning of Liff

This bug was fixed in the package sakura - 3.0.4-2

---------------
sakura (3.0.4-2) unstable; urgency=low

  * debian/patches: Drop all disabled patches and
    close-right-page-on-child-exit.patch, it was also
    applied upstream.

 -- Andrew Starr-Bochicchio <email address hidden> Thu, 03 May 2012 02:20:08 -0400

Hello! Please add Impact/Test case/Regression potential information for this bug as specified by https://wiki.ubuntu.com/StableReleaseUpdates

The upload itself looks great, so just add that and it should be accepted.

Done. Thanks for the reminder.

Hello Andrew, or anyone else affected,

Accepted sakura into precise-proposed. The package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Confirmed the fix

This bug was fixed in the package sakura - 3.0.3-2ubuntu0.1

---------------
sakura (3.0.3-2ubuntu0.1) precise-proposed; urgency=low

  * Drop fix_FTBFS4Hurd.patch; not needed for Ubuntu
    and introduces a bug where opening a new tab does
    not keep the working directory (LP: #977232).
 -- Andrew Starr-Bochicchio <email address hidden> Thu, 03 May 2012 02:43:19 -0400