Re: [lmi] PATCH: Switch to using std::filesystem

[Greg Chicares]

On 4/28/21 12:34 PM, Vadim Zeitlin wrote:
> On Wed, 28 Apr 2021 12:17:49 +0000 Greg Chicares <gchicares@sbcglobal.net> 
> wrote:
> 
> GC> On 4/26/21 9:53 PM, Vadim Zeitlin wrote:
> GC> > 
> GC> >  The branch "std-fs" of https://github.com/vadz/lmi.git repository
> GC> I get the following unit-test error for both msw builds:
> GC> 
> GC> Running path_utility_test:
> GC> 
> GC> ???? test failed:   '...-20210428T115328Z.' == '....'
> GC> [file /opt/lmi/src/lmi/path_utility_test.cpp, line 302]
> GC> 
> GC> ???? 1 test errors detected; 64 tests succeeded
> 
>  Very curious. Is this using Wine and, if so, which version? It must be
> Debian Wine 5.0.3, right?

Yes.

> In any case, it must be different from
> wine-stable_6.0.0~focal-1_amd64.deb used by the CI builds because the test
> does pass there.

Curious indeed.

> GC> This affects unit tests with or without the "safe" standard library.
> 
>  I admit I didn't try with the safe standard library, but I don't see how
> could it affect this.

It shouldn't. But wouldn't the CI tests include "safe" unit tests,
as 'nychthemeral_test.sh' does?

>  I strongly suspect it's a bug in Wine, but it's not really clear to me
> what should we do even if we confirm that this is the case: we don't want
> to "fix" the test for Wine if it means breaking it for native MSW and we
> don't want to just disable this test for Wine entirely, as this is what is
> used in the CI builds

I just pushed a commit that is effective though brutal:
  a0ad50975..26dd8aa5c  master -> master
Feel free to refine it, if you're so disposed, e.g. to
enable this test for "good" wine versions only.

> (we could have a CI build running under native MSW
> too, but this would require another effort... it would probably be simpler
> to have a separate workflow taking the binaries produced by this build and
> running the tests under native MSW, I wonder if I should do this).

I dunno; I'll leave that question to your judgement.

>  Would you consider adding https://dl.winehq.org/wine-builds/debian/ to
> your list of repositories (like .github/workflows/ci.yml does for Ubuntu)
> and use the version from there, rather than from Debian? Because I think
> (and can test it to confirm) that using the latest version should fix this
> problem.

No. That would trade the wine defects in debian-testing
for those in wine's own rolling releases, which could be
a perpetual source of woe. And I strongly prefer to keep
my debian system debian-only.

That means I won't be testing in the same environment as
github's CI, but that's a good thing: diversity in
testing is beneficial.