Skip to content

RFC musl multilib: loader path and headers

[email protected] openembedded-core 1 version New

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 RFC Latest4 patches 0 repliesNew

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/4gcc: point the musl dynamic linker at nonarch_base_libdirmeta/classes-recipe/cross-canadian.bbclass meta/classes-recipe/cross.bbclass meta/classes-recipe/crosssdk.bbclass +5 more+40 -270 replies
2/4musl: wrap the bits headers for multilibmeta/recipes-core/musl/musl_git.bb+8 -10 replies
3/4multilib_header_wrapper.h: support muslscripts/multilib_header_wrapper.h+18 -40 replies
4/4multilib_header: wrap musl headers for multilib buildsmeta/classes-recipe/multilib_header.bbclass+1 -10 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.

Read the v1 thread

Cover letter [RFC PATCH 0/4] musl multilib: loader path and headers

From: "mark.yang" <[email protected]>

This series fixes the loader path and the headers for musl multilib ([YOCTO #13122]). It is not complete, so it keeps the musl multilib guard (4da308d37a).

What it fixes:
- gcc wrote the 64-bit musl loader as /usr/lib64/ld-musl-*.so.1, but musl installs it in nonarch_base_libdir. 64-bit programs could not start, and meson's sanity check failed.
- The multilib header wrapper used bits/wordsize.h, which glibc has and musl does not, so oe_multilib_header did not support musl. On musl the wrapper now checks which of bits/alltypes-32.h and -64.h the sysroot has, and uses the pointer size when it has both.

Not done yet:
- ldd: both variants install /usr/bin/ldd.
- SWIG recipes that read wrapped headers need -D__MHWORDSIZE. python3-m2crypto (meta-python) is the one I found.

To try it, revert 4da308d37a ("musl: Show error when used with multilibs"), which added the guard in musl.inc.

With it reverted, qemuarm64 + lib32 armv7ahf and qemux86-64 + lib32 x86 (usrmerge, ipk) build core-image-minimal and the lib32 SDK. The gcc change was also built without usrmerge, and with buildtools-extended-tarball testsdk.

I am sending this as an RFC because I am not sure this is the right way to do musl multilib.

mark.yang (4):
gcc: point the musl dynamic linker at nonarch_base_libdir
musl: wrap the bits headers for multilib multilib_header_wrapper.h: support musl
multilib_header: wrap musl headers for multilib builds

 meta/classes-recipe/cross-canadian.bbclass    |  1 +
 meta/classes-recipe/cross.bbclass             |  1 +
 meta/classes-recipe/crosssdk.bbclass          |  1 +
 meta/classes-recipe/multilib_header.bbclass   |  2 +-
 meta/recipes-core/musl/musl_git.bb            |  9 ++++-
 meta/recipes-devtools/gcc/gcc-common.inc      |  1 +
 .../gcc/gcc-configure-common.inc              |  2 +
 .../gcc/gcc-multilib-config.inc               | 20 ++++++----
 ...AMIC_LINKER-and-UCLIBC_DYNAMIC_LINKE.patch | 39 ++++++++++---------
 ...s-fix-v4bx-to-linker-to-support-EABI.patch |  2 +-
 scripts/multilib_header_wrapper.h             | 22 +++++++++--
 11 files changed, 67 insertions(+), 33 deletions(-)