Re: [lmi] Strange new GUI tests failures in the CI builds

[Vadim Zeitlin]

On Sat, 5 Mar 2022 21:33:58 +0000 Greg Chicares <gchicares@sbcglobal.net> wrote:

GC> On 2022-03-05 21:17, Vadim Zeitlin wrote:
GC> [...]
GC> >  weird assert
GC> > failures in the latest GUI tests runs in the CI builds. Copying the log
GC> > from https://github.com/let-me-illustrate/lmi/runs/5428529518
GC> 
GC> I think that page says this succeeds for 64-bit msw, and
GC> fails only for 32-bit msw.

 No, sorry, it's the converse: 64 bit build succeeds, 32 bit build fails
(see the tail end of my initial message -- the one I wrote that you
shouldn't read).

GC> > input_sequences: ERROR (Assertion failure: strftime() failed [file 
/__w/lmi/lmi/third_party/wx/src/common/datetime.cpp, line 296, in 
wxCallStrftime()].)
GC> 
GC> Doesn't that suggest that the new version of 'wine' has
GC> introduced a defect in the strftime() provided by its
GC> C RTL?

 This was my first hypothesis because Wine version seems to be the only
thing which changed between the last successful build and the failing
builds from today, but as I can't reproduce this locally, I can neither
confirm nor debug it right now. Examining changes to dlls/msvcrt/time.c
between wine-5.0.3 and wine-6.0.2 shows a somewhat suspicious commit
560ac90ba05 (msvcr100: Use LC_TIME code page when converting strftime
data., 2021-01-04), but I don't see anything really wrong with it and
running the test in a different, or even invalid, locale locally still
doesn't reproduce the problem. I still think it's something about Wine, so
I want to modify the CI builds to let me download the binaries produced by
them and test them locally to check this.

 BTW, do you think it could be useful to make these binaries (or the entire
contents of the installation directory) available more permanently? I.e.
should I do this change only on my local branch or would it make sense to
merge it into master once it works?

 Thanks,
VZ

pgpedivaZKse_.pgp
Description: PGP signature