musl 0.8.8 released; all users should upgrade or patch

musl 0.8.8 released; all users should upgrade or patch

[Rich Felker]

Hi everyone,

I've released musl 0.8.8 including a fix for the buffer overflow issue
discovered yesterday. Release summary:

    Fix for critical buffer overflow vulnerability in fprintf with
    unbuffered files. Major math library correctness and performance
    improvements, including x86 assembly. New floating point
    parser/converter with correct rounding, new scanf that corrects
    many corner-case errors, and new wcstod/f/ld (previously missing).
    Various compatibility improvements and small bug fixes.

    http://www.etalabs.net/musl/releases/musl-0.8.8.tar.gz

With this release, musl seems to have complete interface-level
coverage of ISO C99 and POSIX 2008. There are still some features
missing, however; for example, the 'm' modifier to scanf (for auto
malloc) does not work. Further development in the 0.8 series will aim
to fix these last deficiencies, bring the wctype functions up to date
with current Unicode, and further optimize, clean up, and simplify the
math, float parsing, and scanf code. I'd also like to get the
in-progress _BSD_SOURCE feature test macro patchset finished and
integrated. If all goes well, we might just have one or two more
releases in this series before 0.9 begins.

Looking towards 0.9, at that point I'd like to be optimizing the regex
engine (TRE), adding dlopen support for static-linked programs,
porting to at least one or two more important targets (like mips),
adding a few more important encodings to iconv, and writing
documentation for musl.

And probably some other things I forgot.

Just looked back at the list archives and my original goals for 1.0,
and it seems we're nearly there, and already exceeding the goals in
some areas like application compatibility. A big thanks goes out to
everyone who's been testing and reporting the issues you run into
building apps!

Rich

Re: musl 0.8.8 released; all users should upgrade or patch

[Isaac Dunham]

On Wed, 18 Apr 2012 05:07:56 -0400
Rich Felker <dalias@aerifal.cx> wrote:

> Hi everyone,
> 
> I've released musl 0.8.8 including a fix for the buffer overflow issue
> discovered yesterday. Release summary:
> 
>     Fix for critical buffer overflow vulnerability in fprintf with
>     unbuffered files. Major math library correctness and performance
>     improvements, including x86 assembly. New floating point
>     parser/converter with correct rounding, new scanf that corrects
>     many corner-case errors, and new wcstod/f/ld (previously missing).
>     Various compatibility improvements and small bug fixes.
> 
>     http://www.etalabs.net/musl/releases/musl-0.8.8.tar.gz
> 
> With this release, musl seems to have complete interface-level
> coverage of ISO C99 and POSIX 2008. There are still some features
I forget what they were, but ISTR seeing one or two missing functions
in one of the headers I worked on.

> math, float parsing, and scanf code. I'd also like to get the
> in-progress _BSD_SOURCE feature test macro patchset finished and
> integrated. If all goes well, we might just have one or two more
> releases in this series before 0.9 begins.
What I'd had in mind was to have _BSD_SOURCE ready "before 0.9.0" (when
you were talking about having the license change).
I haven't made much progress on it lately, and have been going through
header by header (there are no patches I haven't sent left).  If anyone
wants to work on the other headers, they're welcome to do so.  A header
should take less than 3 hours to go over, from my experience.

> Just looked back at the list archives and my original goals for 1.0,
> and it seems we're nearly there, and already exceeding the goals in
> some areas like application compatibility. A big thanks goes out to
> everyone who's been testing and reporting the issues you run into
> building apps!

You're welcome, and thank you for developing musl!

Isaac Dunham

Re: musl 0.8.8 released

[Isaac Dunham]

On Wed, 18 Apr 2012 19:13:28 -0700
Isaac Dunham <idunham@lavabit.com> wrote:

> On Wed, 18 Apr 2012 05:07:56 -0400
> Rich Felker <dalias@aerifal.cx> wrote:
> 
> > With this release, musl seems to have complete interface-level
> > coverage of ISO C99 and POSIX 2008. There are still some features

> I forget what they were, but ISTR seeing one or two missing functions
> in one of the headers I worked on.
Ignore the above claim--looking back over the headers, I guess I was
wrong.

Isaac Dunham

Re: musl 0.8.8 released

[Szabolcs Nagy]

* Isaac Dunham <idunham@lavabit.com> [2012-04-18 22:21:57 -0700]:
> Isaac Dunham <idunham@lavabit.com> wrote:
> 
> > On Wed, 18 Apr 2012 05:07:56 -0400
> > Rich Felker <dalias@aerifal.cx> wrote:
> > 
> > > With this release, musl seems to have complete interface-level
> > > coverage of ISO C99 and POSIX 2008. There are still some features
> 
> > I forget what they were, but ISTR seeing one or two missing functions
> > in one of the headers I worked on.
> Ignore the above claim--looking back over the headers, I guess I was
> wrong.
> 

btw there are approximate api coverage tables
(generated using ctags and nm vs posix functions and c99 annex b)

http://repo.or.cz/w/musl-tools.git/blob_plain/HEAD:/tab_posix.html
http://repo.or.cz/w/musl-tools.git/blob_plain/HEAD:/c99.html

c99 is really completely covered
(wchar_t is misreported, imaginary is compiler issue, fma macros are
arch dependent)

Re: musl 0.8.8 released; all users should upgrade or patch

[Rich Felker]

Hi,
There's a nasty problem that slipped through in the new strtol family
code; numbers with leading spaces are rejected. Greatest apologies; it
happened from using an old version of my draft code when I merged
rather than the correct version with the space-skip like strtod did.
I've fixed them both now to include the space-skipping code directly
in __intscan and __floatscan instead of having the wrapper be
responsible, but the wide wrappers still need to be responsible
because they have to use iswspace rather than isspace.

For today, please just use git or manually apply the last few patches
from git to 0.8.8. I'll release 0.8.9 with this issue fixed, and any
fixes for other bugs that pop up in the meantime, tonight.

Rich