Musl Libc Support Lands In Mainline GCC

[phoronix]

Phoronix: Musl Libc Support Lands In Mainline GCC

Musl has long aimed at being a lightweight, simple, free, and correct libc library. However, hindering its adoption has been out-of-tree patches required against GCC for supporting the Musl C library. Fortunately, Musl support has now been merged into GCC...

http://www.phoronix.com/scan.php?page=news_item&px=Musl-Libc-GCC-Support

[xeekei]

I've heard good things about musl. Unlike glibc, in fact.

[computerquip]

Musl is primarily intended for embedded/mobile devices

I don't think this is true actually, based off their introduction page and FAQ which mentions nothing of the sort.

[gens]

I've heard good things about musl. Unlike glibc, in fact.

they bout have their advantages and disadvantages

[pal666]

[/FONT][/COLOR]I don't think this is true actually, based off their introduction page and FAQ which mentions nothing of the sort.

it is true actually, they adveritse less memory usage as their main advantage(with less (working)features being main disadvantage)

[xeekei]

they bout have their advantages and disadvantages

Yeah, the established defaults always get extra crap thrown at them for some reason.

[computerquip]

it is true actually, they adveritse less memory usage as their main advantage(with less (working)features being main disadvantage)

How does that imply that it's intended for use on embedded?

[mlau]

How does that imply that it's intended for use on embedded?

Most truly embedded systems (i.e. some board tucked away in some industrial machinery) don't need even a quarter of what glibc offers. Glibc is huge (1.5MB for example on MIPS32, +900kB for libm), and some of these systems come only with 4-8MB of flash so glibc is a no-go (QNX has a neat option to strip the libc (and other system libraries) of all symbols which are unused in the final image, which sometimes saved 1-2MB).

[computerquip]

Most truly embedded systems (i.e. some board tucked away in some industrial machinery) don't need even a quarter of what glibc offers. Glibc is huge (1.5MB for example on MIPS32, +900kB for libm), and some of these systems come only with 4-8MB of flash so glibc is a no-go (QNX has a neat option to strip the libc (and other system libraries) of all symbols which are unused in the final image, which sometimes saved 1-2MB).

My point still stands.

[ncopa]

I have used musl libc for more than a year now and I must say that I really love it. They do a beautiful job to do things technically correct. They prefer to drop the "feature" when it is not possible to solve the problem in a technically correct way. They do an excellent job in sticking to standards and try force avoid special handling per platform. The result is that code quality is improved many places. Even glibc gets better due to the work on musl libc.

An example of mesa code quality improvement as a result of musl libc:

http://cgit.freedesktop.org/mesa/mesa/commit/src/mesa/main/querymatrix.c?id=2e2087a9ebe1f047dff217ab084972935eb52d1b