musl 0.7.10 released

musl 0.7.10 released

[Rich Felker]

Support for numeric IPv6 address parsing has been added. Compiling
musl with PCC should now succeed and result in a working library. The
heap sort implementation of qsort has been replaced by smooth sort,
yielding nearly-linear run time on arrays which are already
mostly-sorted. Various bugs including printf floating point rounding
and scanf EOF handling have been fixed, as well as minor prototype
issues in the public headers and an issue where variadic functions
were incorrectly inlined on i386 with recent gcc versions.

http://www.etalabs.net/musl/releases/musl-0.7.10.tar.gz

Re: musl 0.7.10 released

[Szabolcs Nagy]

* Rich Felker <dalias@aerifal.cx> [2011-05-17 21:49:47 -0400]:
> Support for numeric IPv6 address parsing has been added. Compiling
> musl with PCC should now succeed and result in a working library. The
> heap sort implementation of qsort has been replaced by smooth sort,
> yielding nearly-linear run time on arrays which are already
> mostly-sorted. Various bugs including printf floating point rounding
> and scanf EOF handling have been fixed, as well as minor prototype
> issues in the public headers and an issue where variadic functions
> were incorrectly inlined on i386 with recent gcc versions.
> 
> http://www.etalabs.net/musl/releases/musl-0.7.10.tar.gz

nice

i see compiler workaround
- pcc preprocessor bug with recursive macro expansion
in WHATSNEW

have you reported the issue to the pcc developers?
or can you show me an example where pcc is incorrect?

Re: musl 0.7.10 released

[Rich Felker]

On Wed, May 18, 2011 at 01:49:47PM +0200, Szabolcs Nagy wrote:
> nice
> 
> i see compiler workaround

Yes. Fortunately it's not so ugly and it might even make the code more
readable to people not familiat with the intricacies of the C
preprocessor.

> - pcc preprocessor bug with recursive macro expansion
> in WHATSNEW
> 
> have you reported the issue to the pcc developers?

No. I was planning to but haven't gotten around to it. If you or
someone else would be willing to handle it I'd appreciate it.

> or can you show me an example where pcc is incorrect?

See src/internal/syscall.h

I haven't tried reproducing it with a simpler test case, but the basic
issue is that the preprocessor does not suppress recursive macro
expansion. The __syscall_cp() macro expands to __syscall_cpN (where
N=0..6 depending on the number of arguments), which is another macro
which expands to __syscall_cp(...). Per the C standard (since original
ANSI C), since this instance of __syscall_cp was generated from
expanding the __syscall_cp macro earlier, it's not subject to being
expanded as a macro again, but pcc tries to expand it anyway and gets
in an infinite loop.

Rich

Re: musl 0.7.10 released

[Szabolcs Nagy]

* Rich Felker <dalias@aerifal.cx> [2011-05-18 08:32:54 -0400]:
> > have you reported the issue to the pcc developers?
> 
> No. I was planning to but haven't gotten around to it. If you or
> someone else would be willing to handle it I'd appreciate it.
> 

ok i reported the issue (or at least something related)
http://marc.info/?l=pcc-list&m=130573226614698

> > or can you show me an example where pcc is incorrect?
> 
> See src/internal/syscall.h
> 
> I haven't tried reproducing it with a simpler test case, but the basic
> issue is that the preprocessor does not suppress recursive macro

thanks