[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <9f570c85-574d-4944-9ff6-55fcebc3a155@app.fastmail.com>
Date: Fri, 12 Dec 2025 04:02:00 +0100
From: Alex Rønne Petersen <alex@...xrp.com>
To: "Rich Felker" <dalias@...c.org>
Cc: musl@...ts.openwall.com
Subject: Re: [PATCH] powerpc64: add nop after bl to start_c
On Fri, Dec 12, 2025, at 03:30, Rich Felker wrote:
> On Thu, Dec 11, 2025 at 07:00:17PM +0100, Alex Rønne Petersen wrote:
>> To avoid errors like this:
>>
>> error: ld.lld: /home/ci/.cache/zig/o/ac8c1808c4a1f20ad8a0b179b5207ee7/crt1.o:(.text+0x24): call to save__start_c lacks nop, can't restore toc
>> ---
>
> Just a linker error message isn't sufficient justification for a
> change; it very well may be a linker bug, possibly a newly-introduced
> one. Can you explain what this message means, if/why you think it's
> valid, and which particular lld versions it's happening with?
For the second patch (__dlsym), I'm fairly sure LLD is wrong. This one I'm less sure about, though I would say it's *probably* also wrong here. The LLD version is 21.1.8, and the relevant code is here: https://github.com/llvm/llvm-project/blob/96b6594c901d0d92eec12c9e06388d7011049f69/lld/ELF/Arch/PPC64.cpp#L1726-L1751
I would be surprised if this is a new bug; I think it's more likely that it's always been there and nobody has hit it before because you seem to need rather specific conditions to do so. We do quite often manage to expose bugs relating to branch range, relocation range, range thunks, etc due to Zig's compilation model, which just dumps an entire library/executable worth of code into a single object file.
(See also my reply on the second patch.)
>
> Rich
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
