Explore what parts of musl libc can be easily imported in to the APE libs

[staalmannen]

• musl math library - easy ports, architecture-independent code
• musl math library - functions that depend on features not supported by the compiler
• musl complex
• fenv - easy ports, architecture-independent code
• fenv - architecture-dependent code
• musl search
• musl regex
• musl wchar functions
• musl string functions
• musl iconv
• musl libintl
• musl stdio

In a recent porting attempt, I found myself importing several math functions from musl libc to /sys/src/ape/lib/ap/math. They were pretty pain-free and very enjoyable code. If possible I would prefer those implementations to gnulib whenever possible. The gnulib idea is great, but the headers and code are definitely not intended to be read or edited by a human.

The idea is thus to add as much as possible of musl libc to the APE libs. The more functions that are already in libc, the less of gnulib is needed to build various applications.

[staalmannen]

First integration of some math functions from musl libc to APE libraries
9b76ac4

[staalmannen]

A big move towards this aim. All the math functions (not architecture-dependent) imported
5a636d7
The ones that do not build are commented out in sys/src/ape/lib/ap/math/mkfile
One major issue is to convert the 0x(hexadecimal)p(number) type numbers which pcc does not understand.

[staalmannen]

Architecture-independent parts of fenv has been imported
7671269

[staalmannen]

Several extra parts of the musl math library now successfully compiles after the compiler was patched to enable hex float numbers.

f204b71

[staalmannen]

Search module from musl libc imported

8888269

[staalmannen]

46d4ada

This commit represents a huge reorganization where libap, libbsd, libutf, libnet and libv merge and sys/src/ape/lib/ap is now organized like the src directory in musl. From now on I will always try to default on musl as external source for missing libc functionality.

Things are most likely NOT building right now. What I hope for is "short term pain for long term gain"

[staalmannen]

the merged libap builds. Some newly imported musl functions had to be disabled for now, will fix later. Will have to go through the headers to expose all new functions that have been imported. With the new organization, it should be far easier in the future to identify missing bits and copy them from musl.

820b806

at the moment of this commit, the build failed at the libpcre stage since we are now using musl regex + regex header and it got confused. Next thing to fix.

[staalmannen]

The reorganization and rebuild encountered a couple of more issues unfortunately. As soon as I can rebuild from scratch again, I will probably make an intermediate release.

[staalmannen]

After the successful build and intermediate release, there might be some additional disruptive reorganizations. One thing might be to rip out the APE stdio core and replace it with musl. The reason for this is that we now, for example, have to have 2 different FILE structs, which is not very nice. The musl stdio and FILE struct is far more compatible with 3rd party applications.

[staalmannen]

The entire (architecture-independent) musl libm compile thanks to some compiler hacks

ff00925

[staalmannen]

For fenv architecture dependent code: the .s files I converted with the "as" tool are not working. I need a fix for as

[staalmannen]

migrated to musl stdio, and it seems to be working
bdeec93

architecture-dependent fenv .s file conversion was also already done, Only tested on amd64
20778a7

musl iconv has now replaced the free standing GNU libiconv
bdeec93