[PATCH] msvcrt: Fix fscanf return when EOF is immediately after an end of line.

[Erich E. Hoover]

Ah, well - glad I asked :)  How much specificity do you want in
figuring out where this got fixed? (since there are no tests before
msvcr90)

Best,
Erich

On Fri, Oct 18, 2019 at 1:22 PM Piotr Caban <piotr.caban at gmail.com> wrote:
>
> Hi,
>
> The test is not failing on msvcr120 only because it’s skipped on all machines.
>
> The tests are linked agains msvcrt.dll, you need to load the function dynamically.
>
> Thanks,
> Piotr
>
> > On 18 Oct 2019, at 21:15, Erich E. Hoover <erich.e.hoover at gmail.com> wrote:
> >
> > Wow, thanks Piotr!
> >
> > Are you sure you were testing against msvcr90?  I just ran a couple
> > tries against the testbot and I see it doing what we expect on
> > msvcr120, but msvcr90 it's giving 0 (
> > https://testbot.winehq.org/JobDetails.pl?Key=57997 ).  I'll go through
> > and establish exactly what version they fixed this in, but I want to
> > make sure that there isn't something weird going on with msvcr90.
> >
> > Best,
> > Erich
> >
> > On Fri, Oct 18, 2019 at 12:45 PM Piotr Caban <piotr.caban at gmail.com> wrote:
> >>
> >> Hi Erich,
> >>
> >> I've done some testing and here's what I found:
> >>  - swscanf in msvcrt returns 0 as you have described
> >>  - swscanf in msvcr90 returns WEOF
> >>  - __stdio_common_vswscanf in ucrtbase returns WEOF
> >>
> >> It looks like there's a bug/backward compatibility code in native
> >> msvcrt.dll. It would be good to do some more testing on different
> >> *wscanf functions in msvcrt.
> >>
> >> I guess that the way to go is to:
> >>  - add swscanf tests in msvcrt (you can use L"" there)
> >>  - add swscanf tests in one of the newer dlls
> >>  - make the implementation change specific to newer versions of C-runtime
> >>
> >> Hope it helps,
> >> Piotr
> >>
> >> On 10/18/19 5:40 PM, Erich E. Hoover wrote:
> >>> Hi Piotr,
> >>>
> >>> I just ran the widecharstring case on the testbot (
> >>> https://testbot.winehq.org/JobDetails.pl?Key=57989 ) and that seems to
> >>> return 0 instead of WEOF...  Are you alright with just adding this
> >>> test to demonstrate that it doesn't work the same way?  Failing test:
> >>> ===
> >>>     buffer[0] = ' ';
> >>>     buffer[1] = '\t';
> >>>     buffer[2] = '\n';
> >>>     buffer[3] = '\n';
> >>>     buffer[4] = 0;
> >>>     ret = swscanf(buffer, formats, results);
> >>>     ok( ret == (short)WEOF, "ret = %d\n", ret );
> >>> ===
> >>>
> >>> Best,
> >>> Erich
> >>>
> >>> On Fri, Oct 18, 2019 at 3:49 AM Piotr Caban <piotr.caban at gmail.com> wrote:
> >>>>
> >>>> Hi Erich,
> >>>>
> >>>> On 10/17/19 5:41 PM, Erich E. Hoover wrote:
> >>>>> +                    /* if we have reached the EOF and output nothing then report EOF */
> >>>>> +                    if (nch==_EOF_  && rd==0 && st==0) {
> >>>> _UNLOCK_FILE_(file);
> >>>>> +                        return _EOF_RET;
> >>>>> +                    }
> >>>>
> >>>> Please also update the widecharstring case.
> >>>>
> >>>>> +    fp = fopen(file_name, "wb");
> >>>>> +    ok(fp, "fp = %p\n", fp);
> >>>> This cause compilation warning.
> >>>>
> >>>> Thanks,
> >>>> Piotr
> >>
>