Testsuite failures in A05execution inside Jenkins, but not locally (was: zsh 5.0.8.-test-3)

Testsuite failures in A05execution inside Jenkins, but not locally (was: zsh 5.0.8.-test-3)

[Axel Beckert]

Hi Peter,

On Tue, Aug 25, 2015 at 07:48:07PM +0100, Peter Stephenson wrote:
> I've uploaded 5.0.8-test-3 to
> 
> http://www.zsh.org/pub/development/
[...]
> If you have a not very common operating system, please try this out;
> otherwise, 5.1 should be fit to release.

Sorry for my silence this time, I'm currently in that
busy-after-back-from-holidays state.

I managed to import 5.0.8-test-2 into Debian's Zsh packaging git
repository and it build fine locally.

But I had test suite failures when being build inside a Jenkins job.
Doesn't seem to be caused by the zsh updated though as the most recent
5.0.8 build failed at the very same test:

../../Test/A05execution.ztst: starting.
[…]
Running test: Check whether '$pipestatus[]' behaves.
Unable to change MONITOR option
Test successful.
Running test: Check $pipestatus with a known difficult case
Test successful.
Running test: Bug regression: piping a shell construct to an external process may hang
This test takes 5 seconds to fail...
*** /tmp/zsh.ztst.err.17001  Sat Aug 22 01:00:25 2015
--- /tmp/zsh.ztst.terr.17001 Sat Aug 22 01:00:25 2015
***************
*** 0 ****
--- 1,5 ----
+ (eval):printf:4: write error: broken pipe
+ (eval):4: write error: inappropriate ioctl for device
+ cat: write error: Broken pipe
+ hang:printf: write error: broken pipe
+ hang: write error: inappropriate ioctl for device
Test ../../Test/A05execution.ztst failed: error output differs from expected as shown above for:
  { unsetopt MONITOR } 2>/dev/null
  coproc { read -et 5 || { print -u $ZTST_fd KILLED; kill -HUP -$$ } }
  print -u $ZTST_fd 'This test takes 5 seconds to fail...'
  { printf "%d\n" {1..20000} } | ( read -e )
  hang(){ printf "%d\n" {2..20000} | cat }; hang | ( read -e )
  sleep 1 ;: avoid coproc exit race condition
  print -p done
  read -et 6 -p
Was testing: Bug regression: piping a shell construct to an external process may hang
../../Test/A05execution.ztst: test failed.
The following may (or may not) help identifying the cause:
  This test checks for a file descriptor leak that could cause the left
  side of a pipe to block on write after the right side has exited
../../Test/A06assign.ztst: starting.
[…]

Details:
5.0.8-test-2: http://jenkins.grml.org/view/Debian/job/zsh-binaries/201/architecture=amd64/console
5.0.8: http://jenkins.grml.org/view/Debian/job/zsh-binaries/200/architecture=amd64/console

Will try to import and test 5.0.8-test-3 as soon as possible.

		Kind regards, Axel
-- 
/~\  Plain Text Ribbon Campaign                   | Axel Beckert
\ /  Say No to HTML in E-Mail and News            | abe@deuxchevaux.org  (Mail)
 X   See http://www.nonhtmlmail.org/campaign.html | abe@noone.org (Mail+Jabber)
/ \  I love long mails: http://email.is-not-s.ms/ | http://abe.noone.org/ (Web)

Re: Testsuite failures in A05execution inside Jenkins, but not locally (was: zsh 5.0.8.-test-3)

[Bart Schaefer]

On Aug 26,  3:58pm, Axel Beckert wrote:
}
} But I had test suite failures when being build inside a Jenkins job.
} Doesn't seem to be caused by the zsh updated though as the most recent
} 5.0.8 build failed at the very same test:
} 
} + (eval):printf:4: write error: broken pipe
} + (eval):4: write error: inappropriate ioctl for device
} + cat: write error: Broken pipe
} + hang:printf: write error: broken pipe
} + hang: write error: inappropriate ioctl for device
} Was testing: Bug regression: piping a shell construct to an external process may hang


Those broken pipe errors are irrelevant to the test (in fact they're the
expected reason for the printf and cat processes to terminate, but under
normal circumstances that doesn't print an error).  So I think we can
safely throw them away.

Consequently, does this fix it?


diff --git a/Test/A05execution.ztst b/Test/A05execution.ztst
index d5501bb..d9e6409 100644
--- a/Test/A05execution.ztst
+++ b/Test/A05execution.ztst
@@ -207,7 +207,7 @@ F:This similar test was triggering a reproducible failure with pipestatus.
   coproc { read -et 5 || { print -u $ZTST_fd KILLED; kill -HUP -$$ } }
   print -u $ZTST_fd 'This test takes 5 seconds to fail...'
   { printf "%d\n" {1..20000} } | ( read -e )
-  hang(){ printf "%d\n" {2..20000} | cat }; hang | ( read -e )
+  hang(){ printf "%d\n" {2..20000} | cat }; hang 2>/dev/null | ( read -e )
   print -p done
   read -et 6 -p
 0:Bug regression: piping a shell construct to an external process may hang

Re: Testsuite failures in A05execution inside Jenkins, but not locally (was: zsh 5.0.8.-test-3)

[Axel Beckert]

Hi,

On Wed, Aug 26, 2015 at 11:07:02AM -0700, Bart Schaefer wrote:
> } + (eval):printf:4: write error: broken pipe
> } + (eval):4: write error: inappropriate ioctl for device
> } + cat: write error: Broken pipe
> } + hang:printf: write error: broken pipe
> } + hang: write error: inappropriate ioctl for device
> } Was testing: Bug regression: piping a shell construct to an external process may hang
> 
> Those broken pipe errors are irrelevant to the test (in fact they're the
> expected reason for the printf and cat processes to terminate, but under
> normal circumstances that doesn't print an error).  So I think we can
> safely throw them away.
> 
> Consequently, does this fix it?

Not fully, maybe we need more "2>/dev/null"?

Current failure:

Running test: Bug regression: piping a shell construct to an external process may hang
This test takes 5 seconds to fail...
*** /tmp/zsh.ztst.err.6382   Wed Aug 26 21:24:06 2015
--- /tmp/zsh.ztst.terr.6382  Wed Aug 26 21:24:06 2015
***************
*** 0 ****
--- 1,2 ----
+ (eval):printf:4: write error: broken pipe
+ (eval):4: write error: inappropriate ioctl for device
Test ../../Test/A05execution.ztst failed: error output differs from expected as shown above for:
  { unsetopt MONITOR } 2>/dev/null
  coproc { read -et 5 || { print -u $ZTST_fd KILLED; kill -HUP -$$ } }
  print -u $ZTST_fd 'This test takes 5 seconds to fail...'
  { printf "%d\n" {1..20000} } | ( read -e )
  hang(){ printf "%d\n" {2..20000} | cat }; hang 2>/dev/null | ( read -e )
  sleep 1 ;: avoid coproc exit race condition
  print -p done
  read -et 6 -p
Was testing: Bug regression: piping a shell construct to an external process may hang
../../Test/A05execution.ztst: test failed.
The following may (or may not) help identifying the cause:
  This test checks for a file descriptor leak that could cause the left
  side of a pipe to block on write after the right side has exited

Maybe this line needs a "2>/dev/null", too?

>    { printf "%d\n" {1..20000} } | ( read -e )

Will try that.

		Kind regards, Axel
-- 
/~\  Plain Text Ribbon Campaign                   | Axel Beckert
\ /  Say No to HTML in E-Mail and News            | abe@deuxchevaux.org  (Mail)
 X   See http://www.nonhtmlmail.org/campaign.html | abe@noone.org (Mail+Jabber)
/ \  I love long mails: http://email.is-not-s.ms/ | http://abe.noone.org/ (Web)

Re: Testsuite failures in A05execution inside Jenkins, but not locally (was: zsh 5.0.8.-test-3)

[Axel Beckert]

Hi,

On Thu, Aug 27, 2015 at 03:47:29PM +0200, Axel Beckert wrote:
> Maybe this line needs a "2>/dev/null", too?
> 
> >    { printf "%d\n" {1..20000} } | ( read -e )
> 
> Will try that.

This one helped. The essence of the patch now looks like this:

-  { printf "%d\n" {1..20000} } | ( read -e )
-  hang(){ printf "%d\n" {2..20000} | cat }; hang | ( read -e )
+  { printf "%d\n" {1..20000} } 2>/dev/null | ( read -e )
+  hang(){ printf "%d\n" {2..20000} | cat }; hang 2>/dev/null | ( read -e )

Thanks!

		Kind regards, Axel
-- 
/~\  Plain Text Ribbon Campaign                   | Axel Beckert
\ /  Say No to HTML in E-Mail and News            | abe@deuxchevaux.org  (Mail)
 X   See http://www.nonhtmlmail.org/campaign.html | abe@noone.org (Mail+Jabber)
/ \  I love long mails: http://email.is-not-s.ms/ | http://abe.noone.org/ (Web)

Re: Testsuite failures in A05execution inside Jenkins, but not locally (was: zsh 5.0.8.-test-3)

[Bart Schaefer]

On Aug 27,  3:47pm, Axel Beckert wrote:
}
} > Consequently, does this fix it?
} 
} Not fully, maybe we need more "2>/dev/null"?

Yes, I'm sure that's it.  The order of the error messages made me think
it was the { printf | cat } that was generating both errors, but of
course there's more than one printf.

Re: zsh 5.0.8.-test-3

[Axel Beckert]

Hi,

On Tue, Aug 25, 2015 at 07:48:07PM +0100, Peter Stephenson wrote:
> I've uploaded 5.0.8-test-3 to
> 
> http://www.zsh.org/pub/development/

JFTR: Build fine on all(*) Debian build daemons where the build
already started:
https://buildd.debian.org/status/package.php?p=zsh&suite=experimental

(Includes the cherry-picked test suite fixes Bart and me posted to the
list after the 5.0.8-test-3 release: 51f5898d and caaed169)

(*) Except on GNU Hurd where the configure script (!) seems to hang
    forever for quite some builds now (including 5.0.8 where it built
    successfully in the past):
    https://buildd.debian.org/status/logs.php?pkg=zsh&arch=hurd-i386
    https://buildd.debian.org/status/fetch.php?pkg=zsh&arch=hurd-i386&ver=5.0.8-test-3-1&stamp=1440751714
    I'd expect that this is an autoconf bug and no zsh-specific issue.

		Kind regards, Axel
-- 
/~\  Plain Text Ribbon Campaign                   | Axel Beckert
\ /  Say No to HTML in E-Mail and News            | abe@deuxchevaux.org  (Mail)
 X   See http://www.nonhtmlmail.org/campaign.html | abe@noone.org (Mail+Jabber)
/ \  I love long mails: http://email.is-not-s.ms/ | http://abe.noone.org/ (Web)

Re: zsh 5.0.8.-test-3

[Bart Schaefer]

On Aug 28,  1:41pm, Axel Beckert wrote:
}
} 
} (*) Except on GNU Hurd where the configure script (!) seems to hang
}     forever for quite some builds now

checking if named FIFOs work... debian/rules:89: recipe for target
'configure-static-stamp' failed

This is running mkfifo() or mknod() from a compiled C snippet, followed
by attempting to open() the FIFO.

}     I'd expect that this is an autoconf bug and no zsh-specific issue.

Looks like a problem with the environment created by /usr/bin/fakeroot
rather than with autoconf or zsh.