binutils: update to 2.44

[MingcongBai]

Topic Description

• gcc+cross: depend on binutils
  Binutils now comes with support for all targets.
• binutils: update to 2.44
  • Enable a unified build for all targets.
  • Revise sed invocations into patches.
  • Track patches at AOSC-Tracking/binutils-gdb @ aosc/binutils-2_44
    (HEAD: 072457058d06965a634c35c44b6ecc3c62609e56).

Package(s) Affected

• binutils: 2.44
• gcc+cross-amd64: 13.2.0+glibc2.38+6-3
• gcc+cross-arm64: 13.2.0+glibc2.38+6-3
• gcc+cross-loongarch64: 13.2.0+glibc2.38+6-3
• gcc+cross-loongson3: 13.2.0+glibc2.38+6-3
• gcc+cross-ppc64el: 13.2.0+glibc2.38+6-3
• gcc+cross-riscv64: 13.2.0+glibc2.38+6-3

Security Update?

No

Build Order

#buildit binutils gcc+cross

Test Build(s) Done

Primary Architectures

• AMD64 amd64
• AArch64 arm64
• LoongArch 64-bit loongarch64

Secondary Architectures

• Loongson 3 loongson3
• PowerPC 64-bit (Little Endian) ppc64el
• RISC-V 64-bit riscv64

[xry111]

I've some doubts on the patches. -O2 shouldn't have an effect on the preprocessing (unless some header is checking __OPTIMIZE__ but then that header would be completely stupid) and we shouldn't need to raise ELF_COMMONPAGESIZE as long as ELF_MAXPAGESIZE is correct.

LoongArch also has ELF_COMMONPAGESIZE = 0x4000 and ELF_MAXPAGESIZE = 0x10000, and per various reports the page size issues are fixed after @xen0n raised ELF_MAXPAGESIZE (w/o changing ELF_COMMONPAGESIZE).

[MingcongBai]

I've some doubts on the patches. -O2 shouldn't have an effect on the preprocessing (unless some header is checking __OPTIMIZE__ but then that header would be completely stupid) and we shouldn't need to raise ELF_COMMONPAGESIZE as long as ELF_MAXPAGESIZE is correct.

LoongArch also has ELF_COMMONPAGESIZE = 0x4000 and ELF_MAXPAGESIZE = 0x10000, and per various reports the page size issues are fixed after @xen0n raised ELF_MAXPAGESIZE (w/o changing ELF_COMMONPAGESIZE).

You are probably right. These were originally sed routines in the prepare script, which as you know, are now discouraged - they came from Fedora, so it might be a good idea to dig into how they came to that hack (just as a note, a few years back, we had a pretty bad practice of syncing patches and hacks from other distributions without due diligence).