Opened 4 days ago

Last modified 28 hours ago

#5634 new enhancement

binutils-2.43.90 (wait for 2.44.0)

Reported by: Xi Ruoyao Owned by: lfs-book
Priority: normal Milestone: 12.3
Component: Book Version: git
Severity: normal Keywords:
Cc:

Description (last modified by Xi Ruoyao)

Create the ticket early to note:

The hard coded line number in the sed for pass 2 needs to be changed to 6031 from 6009.

Change History (6)

comment:1 by Xi Ruoyao, 4 days ago

Description: modified (diff)

comment:2 by Xi Ruoyao, 4 days ago

Now it seems "only" 10 (instead of 12 in the book) gold tests are failing.

comment:3 by Bruce Dubbs, 3 days ago

This is interesting: ​https://sourceware.org/pipermail/binutils/2025-January/138617.html

Especially the line: "Long term however we are planning to deprecate gold."

in reply to:  3 comment:4 by Xi Ruoyao, 3 days ago

Replying to Bruce Dubbs:

This is interesting: ​https://sourceware.org/pipermail/binutils/2025-January/138617.html

Especially the line: "Long term however we are planning to deprecate gold."

I'm in favor of just building the "binutils without gold" tarball if the ​separation happens. Today if a package really peruses "being linked with a modern linker," it uses lld or mold instead of gold.

Last edited 3 days ago by Xi Ruoyao (previous) (diff)

comment:5 by Xi Ruoyao, 3 days ago

I'm using --disable-gold for the new system to see if there would be any impact from the removal.

comment:6 by Xi Ruoyao, 28 hours ago

100+ LLVM tests fail w/o gold because they have -fuse-ld=gold. After using a sed to remove -fuse-ld=gold 23 of them still fails with something like "cannot resolve R_X86_64_8 against hidden symbol for PIE." Perhaps just some linker behavior differences.

Note: See TracTickets for help on using tickets.