emacs-26, make check failures, mingw64

emacs-26, make check failures, mingw64

[Fabrice Popineau]

[-- Attachment #1: Type: text/plain, Size: 979 bytes --]

I still get a couples of failures while running tests :

- test/src/editfns-tests.log

Test format-time-string-with-zone condition:
    (ert-test-failed
     ((should
       (string-equal
        (format-time-string format look -28800)
        "1972-06-30 15:59:59.999 -0800 (-08)"))
      :form
      (string-equal "1972-06-30 15:59:59.999 -0800 (ZZZ)" "1972-06-30
15:59:59.999 -0800 (-08)")
      :value nil))

It is a new failure for a new test introduced by:
commit 541006c53623cb5fb7dfae475baae5d64fc6e9d0

Clearly, the timezone is wrongly formatted.

- test/lisp/ibuffer-tests.log

Test ibuffer-filter-inclusion-8 condition:
    (file-error "Creating file with prefix" "Invalid argument"
"c:/Users/Fabrice/AppData/Roaming/Local/Temp/ibuf-test8c")
   FAILED  11/19  ibuffer-filter-inclusion-8

I don't see how create-file-buffer in this test can ever create a temp file
ending with
an '*'. But I am probably overlooking something. Does it pass on other
platforms ?

Fabrice

[-- Attachment #2: Type: text/html, Size: 1447 bytes --]

Re: emacs-26, make check failures, mingw64

[Paul Eggert]

On 09/19/2017 12:08 PM, Fabrice Popineau wrote:
> Clearly, the timezone is wrongly formatted.

This is because src/w32.c uses the abbreviation "ZZZ" when a POSIX TZ 
abbreviation is not supported by MS-Windows.

One possible fix would be for src/w32.c to call tzset with a ZZZ...Z 
abbreviation, and then to modify tzname[0] afterwards to have a copy of 
the POSIX-specified abbreviation. src/w32.c could use enough Zs so that 
the tzname[0] created by the MS-Windows library would be long enough and 
could be updated in place rather than worrying about our reallocating it.

If that's too much trouble, another workaround would be to change the 
test case so that it is not run on MS-Windows.

Re: emacs-26, make check failures, mingw64

[Eli Zaretskii]

> From: Fabrice Popineau <fabrice.popineau@gmail.com>
> Date: Tue, 19 Sep 2017 21:08:56 +0200
> 
> I still get a couples of failures while running tests :

Thanks, they should be fixed now.

> Test ibuffer-filter-inclusion-8 condition:
> (file-error "Creating file with prefix" "Invalid argument"
> "c:/Users/Fabrice/AppData/Roaming/Local/Temp/ibuf-test8c")
> FAILED 11/19 ibuffer-filter-inclusion-8
> 
> I don't see how create-file-buffer in this test can ever create a temp file ending with 
> an '*'. But I am probably overlooking something. Does it pass on other platforms ?

Yes, they pass on Posix systems, where a file name can include "*".

Re: emacs-26, make check failures, mingw64

[Eli Zaretskii]

> From: Paul Eggert <eggert@cs.ucla.edu>
> Date: Tue, 19 Sep 2017 12:38:46 -0700
> 
> On 09/19/2017 12:08 PM, Fabrice Popineau wrote:
> > Clearly, the timezone is wrongly formatted.
> 
> This is because src/w32.c uses the abbreviation "ZZZ" when a POSIX TZ 
> abbreviation is not supported by MS-Windows.
> 
> One possible fix would be for src/w32.c to call tzset with a ZZZ...Z 
> abbreviation, and then to modify tzname[0] afterwards to have a copy of 
> the POSIX-specified abbreviation. src/w32.c could use enough Zs so that 
> the tzname[0] created by the MS-Windows library would be long enough and 
> could be updated in place rather than worrying about our reallocating it.
> 
> If that's too much trouble, another workaround would be to change the 
> test case so that it is not run on MS-Windows.

I just tweaked the test to match what Windows returns in this case.
The main purpose of the test, AFAIU, is not to test the correctness of
the time-zone translation, and there's nothing wrong in ZZZ given that
the "-08" thing is not supported on Windows.  The rest of the result
is correct, so making non-trivial changes in tzset/tzname sounds too
much trouble for very little gain: code should not generally be
tweaked to fix failing tests with fictitious values.

Thanks.