[PATCH v2] Get rid of `with_multisrctop`, and `MULTISRCTOP`

[John Ericson]

Hi, ping on this. Is it all good? I would appreciate being able to land these changes if possible.

Per the original, I *was* able to test it after all.

John

On Fri, Aug 22, 2025, at 10:01 PM, John Ericson wrote:
> Thanks, both of you, for taking a look
> 
> On 2025-08-21 14:26, Jeff Johnston wrote:
>> According to your comment below, you don't have it working for multilib newlib. 
> 
> Just to be clear, I couldn't before get a --with-multilib build going with or without my patch. So I didn't know whether it worked, *not* knew that I had broken something.
> 
>> Regarding a test - you can try and build aarch64-none-elf.  That will build a 
>> default newlib and one in an ilp32 multilib directory.
> 
> OK so short story, thanks, that worked! The patch is working
> 
> The full story is my distro (Nixpkgs) 's aarch64-none-elf GCC was built with `--disable-multilib`, so I faked it a bit with more or less:
> 
> diff --git a/config-ml.in b/config-ml.in
> index 56b6abae2..578377785 100644
> --- a/config-ml.in
> +++ b/config-ml.in
> @@ -103,6 +103,8 @@
>  # Note that `host' in this case is GCC's `target'.  Target libraries are
>  # configured for a particular host.
> 
> +print_multilib=`printf '.;\nfoo;@mabi=ilp32'`
> +
>  Makefile=${ac_file-Makefile}
>  ml_config_shell=${CONFIG_SHELL-/bin/sh}
> 
> @@ -203,7 +205,7 @@ fi
>  if [ "${ml_toplevel_p}" = yes ]; then
> 
>  multidirs=
> -for i in `${CC-gcc} --print-multi-lib 2>/dev/null`; do
> +for i in $print_multilib; do
>    dir=`echo $i | sed -e 's/;.*$//'`
>    if [ "${dir}" = "." ]; then
>      true
> @@ -484,7 +486,7 @@ multi-do:
>           srcrootpre=`cd $(srcdir); ${PWD_COMMAND}`/; export srcrootpre; \
>           lib=`echo "$${rootpre}" | sed -e 's,^.*/\([^/][^/]*\)/$$,\1,'`; \
>           compiler="$(CC)"; \
> -         for i in `$${compiler} --print-multi-lib 2>/dev/null`; do \
> +         for i in `printf '.;\nfoo;@mabi=ilp32'`; do \
>             dir=`echo $$i | sed -e 's/;.*$$//'`; \
>             if [ "$${dir}" = "." ]; then \
>               true; \
> @@ -693,7 +695,7 @@ if [ -n "${multidirs}" ] && [ -z "${ml_norecursion}" ]; then
>      fi
> 
>      # find compiler flag corresponding to ${ml_dir}
> -    for i in `${CC-gcc} --print-multi-lib 2>/dev/null`; do
> +    for i in $print_multilib; do
>        dir=`echo $i | sed -e 's/;.*$//'`
>        if [ "${dir}" = "${ml_dir}" ]; then
>          flags=`echo $i | sed -e 's/^[^;]*;//' -e 's/@/ -/g'`
> 
> and it worked!
> 
> $ make
> [...]
> Making all in .
> make[8]: Entering directory '/home/jcericson/src/gnu/newlib-cygwin/build/aarch64-none-elf/foo/libgloss'
> true  DO=all multi-do # make
> make[8]: Leaving directory '/home/jcericson/src/gnu/newlib-cygwin/build/aarch64-none-elf/foo/libgloss'
> make[7]: Leaving directory '/home/jcericson/src/gnu/newlib-cygwin/build/aarch64-none-elf/foo/libgloss'
> make[6]: Leaving directory '/home/jcericson/src/gnu/newlib-cygwin/build/aarch64-none-elf/foo/libgloss'
> + true
> make[5]: Leaving directory '/home/jcericson/src/gnu/newlib-cygwin/build/aarch64-none-elf/libgloss'
> make[4]: Leaving directory '/home/jcericson/src/gnu/newlib-cygwin/build/aarch64-none-elf/libgloss'
> make[3]: Leaving directory '/home/jcericson/src/gnu/newlib-cygwin/build/aarch64-none-elf/libgloss'
> make[2]: Leaving directory '/home/jcericson/src/gnu/newlib-cygwin/build/aarch64-none-elf/libgloss'
> make[1]: Leaving directory '/home/jcericson/src/gnu/newlib-cygwin/build
> 
> $ echo $?
> 0
> 
> And grepping confirms the right things happened:
> 
> $ grep -irl ilp32 aarch64-none-elf | head
> aarch64-none-elf/newlib/Makefile
> aarch64-none-elf/newlib/config.log
> aarch64-none-elf/newlib/targ-include/machine/_fpmath.h
> aarch64-none-elf/newlib/config.status
> aarch64-none-elf/foo/newlib/libc.a
> aarch64-none-elf/foo/newlib/config.status
> aarch64-none-elf/foo/newlib/config.cache
> aarch64-none-elf/foo/newlib/config.log
> aarch64-none-elf/foo/newlib/foo/libm.a
> aarch64-none-elf/foo/newlib/foo/libc.a
> 
> $ grep -irl lp64 aarch64-none-elf | head
> aarch64-none-elf/newlib/libc/libm_a-sf_isinf.o
> aarch64-none-elf/newlib/libc/libm_a-s_isnand.o
> aarch64-none-elf/newlib/libc/libm_a-s_isinfd.o
> aarch64-none-elf/newlib/libc/locale/libc_a-nl_langinfo.o
> aarch64-none-elf/newlib/libc/locale/libc_a-duplocale.o
> aarch64-none-elf/newlib/libc/locale/libc_a-locale.o
> aarch64-none-elf/newlib/libc/locale/libc_a-lmonetary.o
> aarch64-none-elf/newlib/libc/locale/libc_a-timelocal.o
> aarch64-none-elf/newlib/libc/locale/libc_a-freelocale.o
> aarch64-none-elf/newlib/libc/locale/libc_a-localeconv.o
> 
> So I think we can now consider the patch tested! To the extent this is artificial, with me possibly missing the proper libgcc etc. from my --disable-multilib GCC, I don't think that matters with the patch I did.
> 
> On Fri, Aug 22, 2025, at 5:44 AM, Brian Inglis wrote:
>> Check that nothing breaks Mike Frysinger's flattened make structure?
> 
> I wasn't sure what you mean by this?
> 
> $ git log --author 'Mike Frysinger' --grep=akefile
> 
> indicates it might be his making libgloss use Automake and less recursive make? I didn't find any special options to test, but I did build both libgloss and newlib
> 
> $ find aarch64-none-elf/ -maxdepth 2| grep libgloss | head -n2
> aarch64-none-elf/foo/libgloss
> aarch64-none-elf/libgloss
> 
> so I didn't break libgloss either.
> 
> Let me know if this is amount of testing good, or if there is any more information you need.
> 
> Thanks,
> 
> John
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://sourceware.org/pipermail/newlib/attachments/20251013/b0fd44c8/attachment-0001.htm>