Re: [PATCH 08/12] mingw: rely on MSYS2's metadata instead of hard-coding it

Junio C Hamano <[email protected]> Wed, 05 Aug 2026 10:29:49 -0700
Newsgroups org.kernel.vger.git
Message-ID <[email protected]>
"Johannes Schindelin via GitGitGadget" <[email protected]>
writes:

> -                ifeq (CLANGARM64,$(MSYSTEM))
> -			prefix = /clangarm64
> -                else
> -			prefix = /mingw64
> -                endif
> +        ifneq (,$(MSYSTEM))
> +		prefix = $(MINGW_PREFIX)
>          endif

Mental note: if $(MSYSTEM) is not an empty string, we set prefix to
$(MINGW_PREFIX).

> @@ -755,6 +749,10 @@ ifeq ($(uname_S),MINGW)
>  		BASIC_LDFLAGS += -Wl,--dynamicbase
>          endif
>          ifneq (,$(MSYSTEM))
> +                ifeq ($(MINGW_PREFIX),$(filter-out /%,$(MINGW_PREFIX)))
> +			# Override if empty or does not start with a slash
> +			MINGW_PREFIX := /$(shell echo '$(MSYSTEM)' | tr A-Z a-z)
> +                endif

Mental note: MINGW_PREFIX that does not begin with a slash is forced
to begin with a slash.

>  		prefix = $(MINGW_PREFIX)

And that becomes $(prefix).

> diff --git a/meson.build b/meson.build
> index 7073d5844d..6ddc461873 100644
> --- a/meson.build
> +++ b/meson.build
> @@ -1318,7 +1318,6 @@ elif host_machine.system() == 'windows'
>  
>    libgit_c_args += [
>      '-DDETECT_MSYS_TTY',
> -    '-DENSURE_MSYSTEM_IS_SET',
>      '-DNATIVE_CRLF',
>      '-DNOGDI',
>      '-DNO_POSIX_GOODIES',
> @@ -1328,6 +1327,18 @@ elif host_machine.system() == 'windows'
>      '-D__USE_MINGW_ANSI_STDIO=0',
>    ]
>  
> +  msystem = get_option('msystem')
> +  if msystem != ''
> +    mingw_prefix = get_option('mingw_prefix')
> +    if mingw_prefix == ''
> +      mingw_prefix = '/' + msystem.to_lower()
> +    endif
> +    libgit_c_args += [
> +      '-DENSURE_MSYSTEM_IS_SET="' + msystem + '"',
> +      '-DMINGW_PREFIX="' + mingw_prefix + '"'
> +    ]
> +  endif

Lowercase mingw_prefix in Meson world corresponds to MINGW_PREFIX in
Make world, I guess.  -DMINGW_PRFIX gets mingw_prefix which begins
with a slash.

I do not do Windows or Meson, but doesn't this contradict with what
we have in [12/12], part of which says:

diff --git a/config.mak.uname b/config.mak.uname
index 2f7d445eb3..0b63be10b7 100644
--- a/config.mak.uname
+++ b/config.mak.uname
@@ -535,7 +535,9 @@ endif
 		compat/win32/pthread.o compat/win32/syslog.o \
 		compat/win32/trace2_win32_process_info.o \
 		compat/win32/dirent.o
-	COMPAT_CFLAGS = -D__USE_MINGW_ACCESS -DDETECT_MSYS_TTY -DNOGDI -DHAVE_STRING_H -Icompat -Icompat/regex -Icompat/win32 -DSTRIP_EXTENSION=\".exe\"
+	COMPAT_CFLAGS = -D__USE_MINGW_ACCESS -DDETECT_MSYS_TTY \
+		-DENSURE_MSYSTEM_IS_SET="\"$(MSYSTEM)\"" -DMINGW_PREFIX="\"$(patsubst /%,%,$(MINGW_PREFIX))\"" \
+		-DNOGDI -DHAVE_STRING_H -Icompat -Icompat/regex -Icompat/win32 -DSTRIP_EXTENSION=\".exe\"
 	BASIC_LDFLAGS = -IGNORE:4217 -IGNORE:4049 -NOLOGO -ENTRY:wmainCRTStartup -SUBSYSTEM:CONSOLE
 	# invalidcontinue.obj allows Git's source code to close the same file
 	# handle twice, or to access the osfhandle of an already-closed stdout


IOW, -DMINGW_PREFIX passed to the compiler strips leading slash from
$(MINGW_PREFIX).

Isn't it necessary to strip the leading slash from ming_prefix also
on the Meson side?