Tue 16 Nov 2010 05:07:30 PM UTC, comment #12:
> Great, the package finally passes all tests :)
That's good news. Thank you, I'm closing this bug.
|
Tue 16 Nov 2010 04:35:49 PM UTC, comment #11:
Puhh i finally found the reason for the failures by testing and diffing the build logs. As i am doing Debian packages, i used "--prefix=/usr" on configure, but that has some some influence on the lib for readline, which i found to be the reason for the failures. So i solved it by using also "--without-libreadline-prefix".
It's already obvious in the configure output, so the tests will fail on
"checking how to link with libreadline... /usr/lib/libreadline.so"
while
"checking how to link with libreadline... -lreadline"
will work afterwards.
Great, the package finally passes all tests :)
|
Tue 16 Nov 2010 12:11:57 AM UTC, comment #10:
>Might the best would be if you maybe be able do a Ubuntu VM too, so you can reproduce it. Or let me know what i should do next.
My Ubuntu 64-bit build passed without any failures: http://pspp.benpfaff.org/builds/104/ubuntu64/summary.html
Perhaps you could turn on core dumps and pass along your binary and the cores that get dumped.
|
Mon 15 Nov 2010 10:59:29 PM UTC, comment #9:
> Might the best would be if you maybe be able do a Ubuntu VM too, so you can reproduce it.
OK, Ubuntu 10.04 LTS AMD64 download in progress.
|
Mon 15 Nov 2010 10:54:08 PM UTC, comment #8:
Okay i rechecked with current git, only 104 & 105 are left as you said. Thanks for that. But backtracing with GDB didn't work.
I also tried it in a maverick chroot for AMD64 as the toolchain should be nearer to squeeze, but just the same errors.
Might the best would be if you maybe be able do a Ubuntu VM too, so you can reproduce it. Or let me know what i should do next.
(file #22027, file #22028)
|
Mon 15 Nov 2010 07:58:42 AM UTC, comment #7:
Oh great to see your efforts. I rechecked the errors on Ubuntu lucid i386, but i also got failures on 104 105 689.
As for playing Sherlock Homes i am of course willing to do so, but this will follow tomorrow, as today is quite busy.
|
Mon 15 Nov 2010 06:23:24 AM UTC, comment #6:
somehow my previous comment was cut off. Continuing:
Run it under GDB with the command "./libtool --mode=execute gdb --args src/ui/terminal/pspp get.sps". At the GDB prompt run "break xalloc_die" and then "run". PSPP should start and then die with the message "xrealloc: out of virtual memory" and come back to a GDB prompt. Run the GDB command "backtrace" and then paste its output into this bug report.
|
Mon 15 Nov 2010 06:20:59 AM UTC, comment #5:
I also pushed fixes for test #680 and #689. That leaves #104 and #105, so I'm updating the title. I can't reproduce those failures here, so bojo42 are you willing to do some detective work on this one? Start by cutting and pasting the following into a file named "get.sps":
Then, from the PSPP build directory, run it under GDB with the command:
|
Mon 15 Nov 2010 05:43:30 AM UTC, comment #4:
I pushed a fix for the failure of test #92 (only observed on my system, not originally reported as part of this bug).
|
Mon 15 Nov 2010 05:22:59 AM UTC, comment #3:
The failures for tests 286 and 299 are single-digit differences in the last digit of precision, so I pushed a change to the tests that ignores this difference.
|
Mon 15 Nov 2010 05:09:01 AM UTC, comment #2:
The following tests failed in bojo42's 64-bit build:
The following tests failed in my own Debian squeeze 64-bit build:
So 104 and 105 fail only in bojo42's build, 92 fails only in my build, and 286, 299, 680, and 689 fail in both.
|
Mon 15 Nov 2010 04:09:07 AM UTC, comment #1:
Thank you for the report. It seems likely that these failures would occur the same way on any 64-bit build, so I'm in the process of installing a 64-bit VM to try to reproduce them.
|
Wed 10 Nov 2010 12:32:13 PM UTC, original submission:
As i got rejected due to spam filters on -unavailable- i repost my mail via web interface:
In the course of updating the Ubuntu/Debian package to a current development snapshot i had to notice that the testsuite is failing at least on Ubuntu 10.04 AMD64 for current git.
I also included the failures on the perl module, although i care more about the first four failures.
|