Windres with .rc file that references utf8 code page manifest doesn't seem to work

[Costas Argyris]

Please disregard my previous email.

Windres is working just fine - running the program under gdb to test the
result
wasn't a very good idea.

I just forgot to send another followup email to say there is no real
problem - sorry about that.

Costas

On Fri, 28 Apr 2023 at 14:15, Jan Beulich <jbeulich@suse.com> wrote:

> On 01.03.2023 14:36, Costas Argyris via Binutils wrote:
> > Hi
> >
> > I am trying to embed a .manifest file as a resource into a windows
> > executable with windres, but it doesn't seem to work.    Here are the
> steps
> > I am following:
> >
> > 1) Resource .rc file:
> >
> > C:\Users\cargyris\temp>cat utf8.rc
> > 1 24 "utf8.manifest"
> >
> > These are the correct defines for embedding a manifest into an
> executable.
> >
> > 2) It references the .manifest file:
> >
> > C:\Users\cargyris\temp>cat utf8.manifest
> > <?xml version="1.0" encoding="UTF-8" standalone="yes"?>
> > <assembly manifestVersion="1.0" xmlns="urn:schemas-microsoft-com:asm.v1">
> >   <application>
> >     <windowsSettings>
> >       <activeCodePage xmlns="
> > http://schemas.microsoft.com/SMI/2019/WindowsSettings
> > ">UTF-8</activeCodePage>
> >     </windowsSettings>
> >   </application>
> > </assembly>
> >
> > This was taken from the official ms docs:
> >
> >
> https://learn.microsoft.com/en-us/windows/apps/design/globalizing/use-utf8-code-page
> >
> > 3) Compile the .rc file with windres into an object COFF file:
> >
> > C:\Users\cargyris\temp>windres utf8.rc utf8.o
> >
> > 4) Compile a dummy .c file with gcc (mingw-w64):
> >
> > C:\Users\cargyris\temp>gcc -g -O0 -c src.c
> >
> > This creates src.o in the temp dir
> >
> > 5) Link the two object files together:
> >
> > C:\Users\cargyris\temp>gcc src.o utf8.o
> >
> > This creates the executable a.exe
> >
> >
> >
> > Now the problem is that this executable should be working fully in UTF-8
> > since we embedded the appropriate manifest into it, but it's not.    To
> > prove this, pass to it an argument that has a Unicode character and
> inspect
> > how that argument gets picked up by the main function using gdb:
> >
> > C:\Users\cargyris\temp>gdb --args a.exe ﹏\src.c
> > ...
> > (gdb) break main
> > Breakpoint 1 at 0x401564: file src.c, line 5.
> > (gdb) run
> > Starting program: C:\Users\cargyris\temp\a.exe "?\src.c"
> > [New Thread 3692.0x1e44]
> > [New Thread 3692.0x20bc]
> >
> > Thread 1 hit Breakpoint 1, main (argc=2, argv=0x741800) at src.c:5
> > 5           printf("sks\n");
> > (gdb) x/4xb argv[1]
> > 0x741770:       0x3f    0x5c    0x73    0x72
> > (gdb) print argv[1]
> > $1 = 0x741770 "?\\src.c"
> > (gdb)
> >
> >
> > The Unicode character never made it to main.    It was this character:
> >
> > https://www.compart.com/en/unicode/U+FE4F
> >
> > which has a 3-byte long UTF-8 representation:
> >
> > 0xEF 0xB9 0x8F
> >
> > so I was expecting to see those 3 bytes in argv[1] in the main function.
> > Instead, main got the question mark and backslash characters.    It
> doesn't
> > help to switch the Command Prompt code page to UTF-8 (65001) with chcp
> > 65001 just before starting gdb, still the same behavior occurs.
> >
> >
> >
> > The .manifest file I used was taken from the official docs and actually
> > works when embedding it with mt.exe as described there - the program
> > properly takes a UTF-8 char array.
>
> Have you inspected the resulting binaries (or perhaps their dumps via
> whatever suitable inspection tool)? There ought to be some difference, and
> perhaps knowing what's different is more helpful here than observations
> from running the program under gdb. May also be worthwhile to check whether
> mt.exe would actually successfully extract the manifest from the windres-
> produced .exe.
>
> > I was under the impression that embedding the manifest with windres
> should
> > have the same effect as embedding it with ms tools,
>
> Hmm, here you say "embedding", whereas above you said you link the COFF
> object. These are entirely different steps, so what about using windres on
> the .exe instead (which as per the doc ought to be possible)?
>
> Jan
>
> > that is, the final
> > executable should be working fully in UTF-8.    This doesn't seem to be
> > happening though.    Is that a problem with windres, the link phase,
> > something I am doing wrong, or is it just not possible due to
> ms-specifics
> > that are not captured by GNU tools?
> >
> > Thanks,
> > Costas
>
>