Debian Bug report logs - #532570
tar - relies on standard SIGPIPE behaviour

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Bastian Blank <waldi@debian.org>

To: submit@bugs.debian.org

Subject: tar - relies on standard SIGPIPE behaviour

Date: Wed, 10 Jun 2009 11:09:32 +0200

[Message part 1 (text/plain, inline)]

Package: tar
Version: 1.22-1
Severity: grave

tar is not longer able to read tar files with some garbage at the end
under some condition. If it is called from the shell, the resulting
SIGPIPE kills the called bzip2 process and tar swallows the error:

| close(3)                                = 0
| waitpid(18165, [{WIFSIGNALED(s) && WTERMSIG(s) == SIGPIPE}], 0) = 18165
| --- SIGCHLD (Child exited) @ 0 (0) ---
| [...]
| exit_group(0)                           = ?

In the other case I call it via a python script. The python interpreter
collects SIGPIPE by default and the setting is persistent in the child
processes. So the SIGPIPE is now collected by bzip2 and converted
into an error:

| write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = -1 EPIPE (Broken pipe)
| --- SIGPIPE (Broken pipe) @ 0 (0) ---
| [...]
| exit_group(1)                           = ?

| close(3)                                = 0
| waitpid(18105, [{WIFEXITED(s) && WEXITSTATUS(s) == 1}], 0) = 18105
| --- SIGCHLD (Child exited) @ 0 (0) ---
| [...]
| exit_group(2)                           = ?

| bzip2: I/O or other error, bailing out.  Possible reason follows.
| bzip2: Broken pipe
|         Input file = (stdin), output file = (stdout)
| tar: Child returned status 1
| tar: Exiting with failure status due to previous errors

It seems that tar now relies on the behaviour that SIGPIPE kills the
child and therfor needs to properly set it up.

Bastian

-- 
You're too beautiful to ignore.  Too much woman.
		-- Kirk to Yeoman Rand, "The Enemy Within", stardate unknown

[signature.asc (application/pgp-signature, inline)]

Message #10 received at 532570@bugs.debian.org (full text, mbox, reply):

From: Bastian Blank <waldi@debian.org>

To: 532570@bugs.debian.org

Subject: Re: Bug#532570: tar - relies on standard SIGPIPE behaviour

Date: Wed, 10 Jun 2009 12:08:57 +0200

[Message part 1 (text/plain, inline)]

tags 532570 patch
thanks

Patch. I cleaned the autoconf cruft from it, please fix that.

Bastian

[diff (text/plain, attachment)]

[signature.asc (application/pgp-signature, inline)]

Message #15 received at 532570-forwarded@bugs.debian.org (full text, mbox, reply):

From: Bdale Garbee <bdale@gag.com>

To: bug-tar@gnu.org

Cc: 532570-forwarded@bugs.debian.org

Subject: [Fwd: Bug#532570: tar - relies on standard SIGPIPE behaviour]

Date: Wed, 10 Jun 2009 07:31:19 -0600

[Message part 1 (text/plain, inline)]

Please comment on the attached patch from Bastian Blank regarding
SIGPIPE behavior proposed for inclusion in my Debian packaging of tar.

Bdale

[Message part 2 (message/rfc822, inline)]

From: Bastian Blank <waldi@debian.org>

To: submit@bugs.debian.org

Subject: Bug#532570: tar - relies on standard SIGPIPE behaviour

Date: Wed, 10 Jun 2009 11:09:32 +0200

[Message part 3 (text/plain, inline)]

Package: tar
Version: 1.22-1
Severity: grave

tar is not longer able to read tar files with some garbage at the end
under some condition. If it is called from the shell, the resulting
SIGPIPE kills the called bzip2 process and tar swallows the error:

| close(3)                                = 0
| waitpid(18165, [{WIFSIGNALED(s) && WTERMSIG(s) == SIGPIPE}], 0) = 18165
| --- SIGCHLD (Child exited) @ 0 (0) ---
| [...]
| exit_group(0)                           = ?

In the other case I call it via a python script. The python interpreter
collects SIGPIPE by default and the setting is persistent in the child
processes. So the SIGPIPE is now collected by bzip2 and converted
into an error:

| write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = -1 EPIPE (Broken pipe)
| --- SIGPIPE (Broken pipe) @ 0 (0) ---
| [...]
| exit_group(1)                           = ?

| close(3)                                = 0
| waitpid(18105, [{WIFEXITED(s) && WEXITSTATUS(s) == 1}], 0) = 18105
| --- SIGCHLD (Child exited) @ 0 (0) ---
| [...]
| exit_group(2)                           = ?

| bzip2: I/O or other error, bailing out.  Possible reason follows.
| bzip2: Broken pipe
|         Input file = (stdin), output file = (stdout)
| tar: Child returned status 1
| tar: Exiting with failure status due to previous errors

It seems that tar now relies on the behaviour that SIGPIPE kills the
child and therfor needs to properly set it up.

Bastian

-- 
You're too beautiful to ignore.  Too much woman.
		-- Kirk to Yeoman Rand, "The Enemy Within", stardate unknown

[signature.asc (application/pgp-signature, inline)]

[Message part 5 (message/rfc822, inline)]

From: Bastian Blank <waldi@debian.org>

To: 532570@bugs.debian.org

Subject: Bug#532570: tar - relies on standard SIGPIPE behaviour

Date: Wed, 10 Jun 2009 12:08:57 +0200

[Message part 6 (text/plain, inline)]

tags 532570 patch
thanks

Patch. I cleaned the autoconf cruft from it, please fix that.

Bastian

[diff (text/plain, attachment)]

[signature.asc (application/pgp-signature, inline)]

Message #20 received at 532570@bugs.debian.org (full text, mbox, reply):

From: Raphael Hertzog <hertzog@debian.org>

To: Bastian Blank <waldi@debian.org>, 532570@bugs.debian.org

Subject: Re: Bug#532570: tar - relies on standard SIGPIPE behaviour

Date: Fri, 12 Jun 2009 17:51:11 +0200

Hi,

On Wed, 10 Jun 2009, Bastian Blank wrote:
> | close(3)                                = 0
> | waitpid(18165, [{WIFSIGNALED(s) && WTERMSIG(s) == SIGPIPE}], 0) = 18165
> | --- SIGCHLD (Child exited) @ 0 (0) ---
> | [...]
> | exit_group(0)                           = ?
> 
> In the other case I call it via a python script. The python interpreter
> collects SIGPIPE by default and the setting is persistent in the child
> processes. So the SIGPIPE is now collected by bzip2 and converted
> into an error:

We have had this in other packages like git-buildpackage which is also in
python (see http://bugs.debian.org/525411). The problem has been fixed in
the python script and not somewhere else.

tar and any other program (dpkg-source) should be able to rightfully make
the assumption that SIGPIPE is not modified unless it modifies it itself.

In other words, you should fix genorig.py and not ask tar to modify its
behaviour.

See
http://git.debian.org/?p=users/agx/git-buildpackage.git;a=commitdiff;h=3e4b08ed095b12675b8865893274c65782720fe6
for an example of how this got fixed in git-buildpackage.

> It seems that tar now relies on the behaviour that SIGPIPE kills the
> child and therfor needs to properly set it up.

dpkg-source relies on it too.

Cheers,
-- 
Raphaël Hertzog

Contribuez à Debian et gagnez un cahier de l'admin Debian Lenny :
http://www.ouaza.com/wp/2009/03/02/contribuer-a-debian-gagner-un-livre/

Message #25 received at 532570@bugs.debian.org (full text, mbox, reply):

From: Bastian Blank <waldi@debian.org>

To: Raphael Hertzog <hertzog@debian.org>

Cc: 532570@bugs.debian.org

Subject: Re: Bug#532570: tar - relies on standard SIGPIPE behaviour

Date: Fri, 12 Jun 2009 21:44:51 +0200

On Fri, Jun 12, 2009 at 05:51:11PM +0200, Raphael Hertzog wrote:
> On Wed, 10 Jun 2009, Bastian Blank wrote:
> > In the other case I call it via a python script. The python interpreter
> > collects SIGPIPE by default and the setting is persistent in the child
> > processes. So the SIGPIPE is now collected by bzip2 and converted
> > into an error:
> We have had this in other packages like git-buildpackage which is also in
> python (see http://bugs.debian.org/525411). The problem has been fixed in
> the python script and not somewhere else.

Please explain.

> tar and any other program (dpkg-source) should be able to rightfully make
> the assumption that SIGPIPE is not modified unless it modifies it itself.

Please explain where this is specified. You can only rely on
_documented_ behaviour and signal(7) specifies:
| During an execve(2), the dispositions of handled signals are reset to
| the default; the dispositions of ignored signals are left unchanged.

Bastian

-- 
Those who hate and fight must stop themselves -- otherwise it is not stopped.
		-- Spock, "Day of the Dove", stardate unknown

Message #30 received at 532570@bugs.debian.org (full text, mbox, reply):

From: Raphael Hertzog <hertzog@debian.org>

To: Bastian Blank <waldi@debian.org>

Cc: 532570@bugs.debian.org

Subject: Re: Bug#532570: tar - relies on standard SIGPIPE behaviour

Date: Fri, 12 Jun 2009 23:04:14 +0200

On Fri, 12 Jun 2009, Bastian Blank wrote:
> > tar and any other program (dpkg-source) should be able to rightfully make
> > the assumption that SIGPIPE is not modified unless it modifies it itself.
> 
> Please explain where this is specified. You can only rely on
> _documented_ behaviour and signal(7) specifies:
> | During an execve(2), the dispositions of handled signals are reset to
> | the default; the dispositions of ignored signals are left unchanged.

Which means that if you choose to ignore signals, you must ensure that
programs that you execute are fine with this... and you can only do that
if you are also the author/maintainer of those programs. Otherwise you
should not change from the default or you should reinstate the default
value before executing subprocesses.

There is some logic behind this, exactly in the same way that you don't
reset the PATH in all applications, you trust your caller to have kept it
to a sane value...

Cheers,
-- 
Raphaël Hertzog

Contribuez à Debian et gagnez un cahier de l'admin Debian Lenny :
http://www.ouaza.com/wp/2009/03/02/contribuer-a-debian-gagner-un-livre/

Message #35 received at 532570@bugs.debian.org (full text, mbox, reply):

From: Bastian Blank <waldi@debian.org>

To: Raphael Hertzog <hertzog@debian.org>

Cc: 532570@bugs.debian.org

Subject: Re: Bug#532570: tar - relies on standard SIGPIPE behaviour

Date: Sat, 13 Jun 2009 00:08:14 +0200

On Fri, Jun 12, 2009 at 11:04:14PM +0200, Raphael Hertzog wrote:
> There is some logic behind this, exactly in the same way that you don't
> reset the PATH in all applications, you trust your caller to have kept it
> to a sane value...

Different problem. tar relies on exceptional states of its subprocesses,
not normal behaviour.

But anyway, you are not the maintainer of tar and this is a regression.

Bastian

-- 
	"That unit is a woman."
	"A mass of conflicting impulses."
		-- Spock and Nomad, "The Changeling", stardate 3541.9

Message #36 received at 532570-forwarded@bugs.debian.org (full text, mbox, reply):

From: Sergey Poznyakoff <gray@gnu.org.ua>

To: "Bdale Garbee" <bdale@gag.com>

Cc: <bug-tar@gnu.org>, <532570-forwarded@bugs.debian.org>

Subject: Re: [Bug-tar] [Fwd: Bug#532570: tar - relies on standard SIGPIPE behaviour]

Date: Thu, 18 Jun 2009 13:10:17 +0300

Bdale Garbee <bdale@gag.com> ha escrit:

> Please comment on the attached patch from Bastian Blank regarding
> SIGPIPE behavior proposed for inclusion in my Debian packaging of tar.

It is reasonable, but it's better be done in main(). I installed the
following patch:

diff --git a/src/tar.c b/src/tar.c
index dbffc2a..e10b804 100644
--- a/src/tar.c
+++ b/src/tar.c
@@ -2454,10 +2454,10 @@ main (int argc, char **argv)
 
   obstack_init (&argv_stk);
 
-#ifdef SIGCHLD
+  /* Ensure default behavior for some signals */
+  signal (SIGPIPE, SIG_DFL);
   /* System V fork+wait does not work if SIGCHLD is ignored.  */
   signal (SIGCHLD, SIG_DFL);
-#endif
 
   /* Decode options.  */
 
Regards,
Sergey

Message #41 received at 532570@bugs.debian.org (full text, mbox, reply):

From: Bastian Blank <waldi@debian.org>

To: 532570@bugs.debian.org

Subject: NMU diff

Date: Tue, 30 Jun 2009 22:40:33 +0200

[Message part 1 (text/plain, inline)]

Attached is the NMU diff.

Bastian

-- 
It is a human characteristic to love little animals, especially if
they're attractive in some way.
		-- McCoy, "The Trouble with Tribbles", stardate 4525.6

[diff (text/plain, attachment)]

Message #46 received at 532570-close@bugs.debian.org (full text, mbox, reply):

From: Bastian Blank <waldi@debian.org>

To: 532570-close@bugs.debian.org

Subject: Bug#532570: fixed in tar 1.22-1.1

Date: Tue, 30 Jun 2009 21:35:07 +0000

Source: tar
Source-Version: 1.22-1.1

We believe that the bug you reported is fixed in the latest version of
tar, which is due to be installed in the Debian FTP archive:

tar_1.22-1.1.diff.gz
  to pool/main/t/tar/tar_1.22-1.1.diff.gz
tar_1.22-1.1.dsc
  to pool/main/t/tar/tar_1.22-1.1.dsc
tar_1.22-1.1_powerpc.deb
  to pool/main/t/tar/tar_1.22-1.1_powerpc.deb



A summary of the changes between this version and the previous one is
attached.

Thank you for reporting the bug, which will now be closed.  If you
have further comments please address them to 532570@bugs.debian.org,
and the maintainer will reopen the bug report if appropriate.

Debian distribution maintenance software
pp.
Bastian Blank <waldi@debian.org> (supplier of updated tar package)

(This message was generated automatically at their request; if you
believe that there is a problem with it please contact the archive
administrators by mailing ftpmaster@debian.org)


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Format: 1.8
Date: Tue, 30 Jun 2009 22:23:48 +0200
Source: tar
Binary: tar
Architecture: source powerpc
Version: 1.22-1.1
Distribution: unstable
Urgency: low
Maintainer: Bdale Garbee <bdale@gag.com>
Changed-By: Bastian Blank <waldi@debian.org>
Description: 
 tar        - GNU version of the tar archiving utility
Closes: 532570
Changes: 
 tar (1.22-1.1) unstable; urgency=low
 .
   * Non-maintainer upload.
   * Set SIGPIPE to default action, patch from upstream. (closes: #532570)
Checksums-Sha1: 
 3938f17b2e55c09c9ef8f6eb483b456d89e4beac 945 tar_1.22-1.1.dsc
 49d4176b481e15fb0e4f5eb2665fbc3dd3e49346 17793 tar_1.22-1.1.diff.gz
 44406954322d91b03aea8f861d77d0a4dbd6d8d1 851444 tar_1.22-1.1_powerpc.deb
Checksums-Sha256: 
 06580d4474510da0bf26eb2fc456e5efeaf673ba9b9c7940b0b276e9cb99250b 945 tar_1.22-1.1.dsc
 b5c08353216eca57d8e662a7d2afbce0e79fe6af8421f64977a7dc7c7934a9ed 17793 tar_1.22-1.1.diff.gz
 e7a170b339bd9771ab07e2192d789320189aa9289d283b64ffc30d207010b4df 851444 tar_1.22-1.1_powerpc.deb
Files: 
 5e1baed4203b60a9d3f554b26199c4fc 945 utils required tar_1.22-1.1.dsc
 6ef3ccbad51c99960cdd8c5c1481b309 17793 utils required tar_1.22-1.1.diff.gz
 5d18f48ecd3b65ca2a665591fbe60328 851444 utils required tar_1.22-1.1_powerpc.deb

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEYEARECAAYFAkpKeGUACgkQLkAIIn9ODhFa/ACg2zHFhyDvlfbc1dBQVqikRnPA
nXAAn2aqHl1/xbwyxB9SsyOC5d7U+x0T
=CdME
-----END PGP SIGNATURE-----

Message #51 received at 532570@bugs.debian.org (full text, mbox, reply):

From: Bdale Garbee <bdale@gag.com>

To: Bastian Blank <waldi@debian.org>, 532570@bugs.debian.org

Subject: Re: Bug#532570: NMU diff

Date: Wed, 01 Jul 2009 15:43:43 -0300

On Tue, 2009-06-30 at 22:40 +0200, Bastian Blank wrote:
> Attached is the NMU diff.

Thanks!  I have that patch in my git tree already, but hadn't made a new
upload yet...  my bad!  [sigh]

Bdale

Send a report that this bug log contains spam.