GNU bug report logs -
#76381
todo-test-add-and-delete-file fails on Ubuntu 24.10
Previous Next
Full log
View this message in rfc822 format
On Fri, 21 Feb 2025 00:01:48 +0100 Stephen Berman <stephen.berman <at> gmx.net> wrote:
> On Thu, 20 Feb 2025 10:13:24 -0800 Paul Eggert <eggert <at> cs.ucla.edu> wrote:
>
>> On 2025-02-20 05:33, Stephen Berman wrote:
>>> On Mon, 17 Feb 2025 17:14:19 -0800 Paul Eggert <eggert <at> cs.ucla.edu> wrote:
>>>> $ cd test
>>>> $ make lisp/calendar/todo-mode-tests
>>>> make[1]: Entering directory
>>>> '/home/eggert/src/gnu/emacs/static-checking/test'
>>>> GEN lisp/calendar/todo-mode-tests.log
>>>> Running 42 tests (2025-02-17 17:04:52-0800, selector ‘(not (or (tag
>>>> :unstable) (tag :nativecomp)))’)
>>>> Enter a non-existing file name:
>>>>
>>>> ... and the test hangs, waiting for me to type a file name I guess.
>>> I cannot reproduce this, either on the master, emacs-30 or emacs-29
>>> branch (on a GNU/Linux system with ext4 file system). For one thing,
>>> when I run the equivalent of `make lisp/calendar/todo-mode-tests' (I
>>> build out-of-tree) I get:
>>> steve [ ~/build/emacs-master/test ]$ make
>>> ~/src/emacs/emacs-master/test/lisp/calendar/todo-mode-tests
>>> make: *** No rule to make target
>>> '/home/steve/src/emacs/emacs-master/test/lisp/calendar/todo-mode-tests'.
>>> Stop.
>>
>> Thanks for looking into it. I wouldn't expect that command to reproduce the
>> bug, though. I can reproduce the problem at the top level with a simple "make
>> check" though this takes a while as it runs all the tests. I can also
>> reproduce the problem by running
>>
>> cd test
>> make lisp/calendar/todo-mode-tests
>>
>> as shown above.
>>
>> I build in-tree; I don't know if that makes a difference.
>
> FWIW, I also tried running `make lisp/calendar/todo-mode-tests' from the
> test directory of the Emacs source tree, but it also failed with "No
> rule to make target".
>
>> The test reliably fails on Ubuntu 24.10 on a zfs filesystem. It reliably
>> succeeds on on the same host when running on a tmpfs filesystem. So possibly
>> this is related to zfs.
>
> That seems plausible since it's reproducible for you. Yet I wouldn't
> have thought that this kind of error could be cause by some property of
> a file system, though admittedly I have next to no technical knowledge
> of file systems.
>
>> To help debug this I am attaching the trace output of the following command
>> when run in the test directory in both environments:
>>
>> LC_ALL=C USERNAME=test HOME=/tmp/home ENV=/tmp/home/mt LOGNAME=test
>> j=/tmp/home b=/tmp/home USER=test BASH_ENV=/tmp/home/mt
>> PATH=/usr/bin:/usr/sbin strace -f -o /tmp/tr make
>> lisp/calendar/todo-mode-tests
>>
>> where /tmp/home contains a single file, /tmp/home/mt, which is empty.
>>
>> The file zfs.tr.xz is the xz-compressed trace output on zfs, and tmpfs.tr.sz
>> is the trace output on tmpfs. Perhaps that can help diagnose the bug.
>
> Thanks for the traces. I couldn't recognize anything in them that
> indicated a reason for the test failure, but here too I lack the
> necessary knowledge. Hopefully someone who has it will be able to shed
> some light.
This test has previously been reported to fail (bug#58473), though the
bug was closed because failure was not reliably reproducible (so it was
probably not on a zfs filesystem).
Steve Berman
This bug report was last modified 105 days ago.
Previous Next
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.