Re: gettext failing msgconv-3 and lang-c++

[Paul E Reimer]

Hello,

> [19:07 reimer@CSI358562 tmp] cc -liconv -o foo foo.c
> [19:07 reimer@CSI358562 tmp] ./foo
> _LIBICONV_VERSION = 0x10b
> no conversion needed: ret=0, input consumed: 6/6, output produced: 6/6
> entire string: ret=0, input consumed: 5/5, output produced: 6/6
> carefully: ret=0, input consumed: 1/1, output produced: 2/2
The libiconv linked to is the native MacOS/Xcode library.

Thanks,

Paul

On 2/13/24 11:40, Bruno Haible wrote:
> Hi,
>
> Paul E Reimer wrote:
> > mco-test3.out is attached, as well as the console output as
> > gettext-msgconv-3.log.
> Thanks. I need more test from you: Can you please run the attached
> program (you probably need to link it with '-liconv'), and report
> the output?
>
> > > * Regarding the lang-c++ failure:
> > >     Can you please try to
> > >       1. apply the attached math.h.diff,
> > >       2. run "make" again,
> > >       3. $ cd gettext-tools/tests
> > >          $ make check TESTS=lang-c++
> > >     Does this work?
> > For this, I was surprised to see that gettext itself contains three
> > different math.h files
> >
> > [10:24 reimer@CSI358562 gettext-0.22.4] find . -name math.h
> > ./libtextstyle/lib/math.h
> > ./gettext-tools/gnulib-lib/math.h
> > ./gettext-runtime/intl/gnulib-lib/math.h
> >
> > I assume that I should apply the patch to
> > ./gettext-tools/gnulib-lib/math.h .
> Yes, that's the one that matters.
>
> > But this does raise a question as to
> > why are there three?  Running diff on them shows that they are indeed
> > all different, varying by various #if statements.
> Different parts of GNU gettext needs different sets of Gnulib modules.
>
> So, after applying the patch, does "make check TESTS=lang-c++" pass?
>
> Bruno