I assumed a wget musl-libc-1.1.24.tar.gz wouldn't bother anyone, and it could
have been a decent test-case for various reasons.
Please keep in mind i had to fork kernel, libc, and a set of ~500 builds for
TinyCC support, and it's final system-integration of a TinyCC driven OS release
that i focus on, not individual implementation details, while i had to patch
into hundreds of components already.

> 
> I have no idea why arm-tcc would behave as you say because it doesn't.

Anyway, a matter of interpretation of the fact, seems this doesn't work either:

.text
.global  START
.type  START ,%function
START :
        mov r11, #0
        mov r14, #0
        ldr r1, [1f]
        add r1, r15, r1
        mov r0, r13
2:      and r12, r0, #-16
        mov r13, r12
        bl  START_c
.weak _DYNAMIC
.hidden _DYNAMIC
.align 2
1:      .word _DYNAMIC-2b

  $ arm-tcc -c arch/arm/crt_arch.S
  arch/arm/crt_arch.S:7: error: operand expected

Obviously some dynamic address calculation is necessary, and a thing like this
wouldn't do neither, syntactically at least:
  ldr r1, [_DYNAMIC-2b]
  arch/arm/crt_arch.S:7: error: local label '2' not found backward
nor:
  ldr r1, [_DYNAMIC-2f]
  arch/arm/crt_arch.S:7: error: invalid operation with label

I fear i'll have to leave that to anyone very much smarter with ARM mnemonics
than myself. And it's more __asm__() magic hiding inside musl-libc arch/arm,
which is a whole separate project to cope with, which i can't nor won't myself.
Question would be, if arm-tcc __asm__ support deserved and was relevant to any
future release tag made (besides other quirks related to aarch32 involved),
and if at all that was any priority for anyone. I merely considered it for
extended test coverage for TinyCC with a _final_ system integration of a
complete TinyCC driven OS release. And for aarch32 this probably doesn't deserve
any more noise of mine on the mailing-list anyway. If anyone else was capable
to address the issue i am willing to re-iterate over ~500builds linking against
musl-libc with arm-tcc then, if and how much would pass compile-time/run-time.
Otherhwise, sorry.

-- 
Michael Ackermann

Attachment: signature.asc
Description: Digital signature

_______________________________________________
Tinycc-devel mailing list
Tinycc-devel@nongnu.org
https://lists.nongnu.org/mailman/listinfo/tinycc-devel

Reply via email to