Update wasi-libc to d03829489904d38c624f6de9983190f1e5e7c9c5

[jedisct1]

Most notable improvements include realpath and stubs for statvfs, chmod, etc.

Other changes are for WASI 0.2 which is going to be painful to support since 3rd party tools written in Rust may become necessary in addition to LLVM (WebAssembly/wasi-sdk#402)

[jedisct1]

This also removes junk we were shipping but irrelevant to webassembly.

[lin72h]

10KLoc reduce, feel so good

[jedisct1]

Added the wasi-libc libdl shim, so that tools such as py2wasm can use zig cc.

[andrewrk]

man. wasi-libc kinda sucks. how about we provide libc via zig code instead?

[jedisct1]

wasi-libc is horrible. And it's getting worse with the wasi preview 2 bits.

But rewriting it in Zig would be a lot of work.

[alexrp]

@jedisct1 I've rebased this on master locally, do you mind if I force-push to your branch? I'd like to get this in for 0.14.0.

[jedisct1]

Sure, go ahead!

There's been changes in wasi-libc since that PR, but the build system has become a giant spaghetti mess.

[alexrp]

Yeah, I figure this is likely to be our last wasi-libc update, what with all the Rust tooling they're apparently mixing in due to the component model stuff. #2879 is probably going to be the way forward in the long term (unless we develop our own component model tooling).

[alexrp]

@andrewrk you ok with taking this as our (probably) last wasi-libc update? (At least unless/until we develop self-hosted component model tooling.)

[jedisct1]

wasi-libc still supports wasi.0.1 (aka wasi-core, wasip1, etc). So it can still be updated; I don't think the Rust tools are required for it.

[andrewrk]

you ok with taking this as our (probably) last wasi-libc update?

Yes 👍

[andrewrk]

this strikes me as strange... wasi doesn't support dynamic linking, so what in the world is this library?

[andrewrk]

weakvoid*dlopen(constchar*name, intflags)
{
error="dlopen not implemented";
returnNULL;
}
weakvoid*dlsym(void*library, constchar*name)
{
error="dlsym not implemented";
returnNULL;
}

oh.

ok, well, it won't be very difficult to implement this in zig 🙄

[alexrp]

It is just another one of those "emulated" WASI libraries that do nothing so existing software can compile. For whatever reason it doesn't follow the usual naming pattern.

FWIW, these all strike me as silly and I wouldn't be opposed to removing them, but that'd probably be separate work.

[jedisct1]

Most notable improvements include realpath and stubs for statvfs, chmod, etc.

Other changes are for WASI 0.2 which is going to be painful to support since 3rd party tools written in Rust may become necessary in addition to LLVM (WebAssembly/wasi-sdk#402)

[jedisct1]

This also removes junk we were shipping but irrelevant to webassembly.

[lin72h]

10KLoc reduce, feel so good

[jedisct1]

Added the wasi-libc libdl shim, so that tools such as py2wasm can use zig cc.

[andrewrk]

man. wasi-libc kinda sucks. how about we provide libc via zig code instead?

[jedisct1]

wasi-libc is horrible. And it's getting worse with the wasi preview 2 bits.

But rewriting it in Zig would be a lot of work.

[alexrp]

@jedisct1 I've rebased this on master locally, do you mind if I force-push to your branch? I'd like to get this in for 0.14.0.

[jedisct1]

Sure, go ahead!

There's been changes in wasi-libc since that PR, but the build system has become a giant spaghetti mess.

[alexrp]

Yeah, I figure this is likely to be our last wasi-libc update, what with all the Rust tooling they're apparently mixing in due to the component model stuff. #2879 is probably going to be the way forward in the long term (unless we develop our own component model tooling).

[alexrp]

@andrewrk you ok with taking this as our (probably) last wasi-libc update? (At least unless/until we develop self-hosted component model tooling.)

[jedisct1]

wasi-libc still supports wasi.0.1 (aka wasi-core, wasip1, etc). So it can still be updated; I don't think the Rust tools are required for it.

[andrewrk]

you ok with taking this as our (probably) last wasi-libc update?

Yes 👍

[andrewrk]

this strikes me as strange... wasi doesn't support dynamic linking, so what in the world is this library?

[andrewrk]

weakvoid*dlopen(constchar*name, intflags)
{
error="dlopen not implemented";
returnNULL;
}
weakvoid*dlsym(void*library, constchar*name)
{
error="dlsym not implemented";
returnNULL;
}

oh.

ok, well, it won't be very difficult to implement this in zig 🙄

[alexrp]

It is just another one of those "emulated" WASI libraries that do nothing so existing software can compile. For whatever reason it doesn't follow the usual naming pattern.

FWIW, these all strike me as silly and I wouldn't be opposed to removing them, but that'd probably be separate work.

[jedisct1]

Most notable improvements include realpath and stubs for statvfs, chmod, etc.

Other changes are for WASI 0.2 which is going to be painful to support since 3rd party tools written in Rust may become necessary in addition to LLVM (WebAssembly/wasi-sdk#402)

[jedisct1]

This also removes junk we were shipping but irrelevant to webassembly.

[lin72h]

10KLoc reduce, feel so good

[jedisct1]

Added the wasi-libc libdl shim, so that tools such as py2wasm can use zig cc.

[andrewrk]

man. wasi-libc kinda sucks. how about we provide libc via zig code instead?

[jedisct1]

wasi-libc is horrible. And it's getting worse with the wasi preview 2 bits.

But rewriting it in Zig would be a lot of work.

[alexrp]

@jedisct1 I've rebased this on master locally, do you mind if I force-push to your branch? I'd like to get this in for 0.14.0.

[jedisct1]

Sure, go ahead!

There's been changes in wasi-libc since that PR, but the build system has become a giant spaghetti mess.

[alexrp]

Yeah, I figure this is likely to be our last wasi-libc update, what with all the Rust tooling they're apparently mixing in due to the component model stuff. #2879 is probably going to be the way forward in the long term (unless we develop our own component model tooling).

[alexrp]

@andrewrk you ok with taking this as our (probably) last wasi-libc update? (At least unless/until we develop self-hosted component model tooling.)

[jedisct1]

wasi-libc still supports wasi.0.1 (aka wasi-core, wasip1, etc). So it can still be updated; I don't think the Rust tools are required for it.

[andrewrk]

you ok with taking this as our (probably) last wasi-libc update?

Yes 👍

[andrewrk]

this strikes me as strange... wasi doesn't support dynamic linking, so what in the world is this library?

[andrewrk]

weakvoid*dlopen(constchar*name, intflags)
{
error="dlopen not implemented";
returnNULL;
}
weakvoid*dlsym(void*library, constchar*name)
{
error="dlsym not implemented";
returnNULL;
}

oh.

ok, well, it won't be very difficult to implement this in zig 🙄

[alexrp]

It is just another one of those "emulated" WASI libraries that do nothing so existing software can compile. For whatever reason it doesn't follow the usual naming pattern.

FWIW, these all strike me as silly and I wouldn't be opposed to removing them, but that'd probably be separate work.

[jedisct1]

Most notable improvements include realpath and stubs for statvfs, chmod, etc.

Other changes are for WASI 0.2 which is going to be painful to support since 3rd party tools written in Rust may become necessary in addition to LLVM (WebAssembly/wasi-sdk#402)

[jedisct1]

This also removes junk we were shipping but irrelevant to webassembly.

[lin72h]

10KLoc reduce, feel so good

[jedisct1]

Added the wasi-libc libdl shim, so that tools such as py2wasm can use zig cc.

[andrewrk]

man. wasi-libc kinda sucks. how about we provide libc via zig code instead?

[jedisct1]

wasi-libc is horrible. And it's getting worse with the wasi preview 2 bits.

But rewriting it in Zig would be a lot of work.

[alexrp]

@jedisct1 I've rebased this on master locally, do you mind if I force-push to your branch? I'd like to get this in for 0.14.0.

[jedisct1]

Sure, go ahead!

There's been changes in wasi-libc since that PR, but the build system has become a giant spaghetti mess.

[alexrp]

Yeah, I figure this is likely to be our last wasi-libc update, what with all the Rust tooling they're apparently mixing in due to the component model stuff. #2879 is probably going to be the way forward in the long term (unless we develop our own component model tooling).

[alexrp]

@andrewrk you ok with taking this as our (probably) last wasi-libc update? (At least unless/until we develop self-hosted component model tooling.)

[jedisct1]

wasi-libc still supports wasi.0.1 (aka wasi-core, wasip1, etc). So it can still be updated; I don't think the Rust tools are required for it.

[andrewrk]

you ok with taking this as our (probably) last wasi-libc update?

Yes 👍

[andrewrk]

this strikes me as strange... wasi doesn't support dynamic linking, so what in the world is this library?

[andrewrk]

weakvoid*dlopen(constchar*name, intflags)
{
error="dlopen not implemented";
returnNULL;
}
weakvoid*dlsym(void*library, constchar*name)
{
error="dlsym not implemented";
returnNULL;
}

oh.

ok, well, it won't be very difficult to implement this in zig 🙄

[alexrp]

It is just another one of those "emulated" WASI libraries that do nothing so existing software can compile. For whatever reason it doesn't follow the usual naming pattern.

FWIW, these all strike me as silly and I wouldn't be opposed to removing them, but that'd probably be separate work.

[jedisct1]

Most notable improvements include realpath and stubs for statvfs, chmod, etc.

Other changes are for WASI 0.2 which is going to be painful to support since 3rd party tools written in Rust may become necessary in addition to LLVM (WebAssembly/wasi-sdk#402)

[jedisct1]

This also removes junk we were shipping but irrelevant to webassembly.

[lin72h]

10KLoc reduce, feel so good

[jedisct1]

Added the wasi-libc libdl shim, so that tools such as py2wasm can use zig cc.

[andrewrk]

man. wasi-libc kinda sucks. how about we provide libc via zig code instead?

[jedisct1]

wasi-libc is horrible. And it's getting worse with the wasi preview 2 bits.

But rewriting it in Zig would be a lot of work.

[alexrp]

@jedisct1 I've rebased this on master locally, do you mind if I force-push to your branch? I'd like to get this in for 0.14.0.

[jedisct1]

Sure, go ahead!

There's been changes in wasi-libc since that PR, but the build system has become a giant spaghetti mess.

[alexrp]

Yeah, I figure this is likely to be our last wasi-libc update, what with all the Rust tooling they're apparently mixing in due to the component model stuff. #2879 is probably going to be the way forward in the long term (unless we develop our own component model tooling).

[alexrp]

@andrewrk you ok with taking this as our (probably) last wasi-libc update? (At least unless/until we develop self-hosted component model tooling.)

[jedisct1]

wasi-libc still supports wasi.0.1 (aka wasi-core, wasip1, etc). So it can still be updated; I don't think the Rust tools are required for it.

[andrewrk]

you ok with taking this as our (probably) last wasi-libc update?

Yes 👍

[andrewrk]

this strikes me as strange... wasi doesn't support dynamic linking, so what in the world is this library?

[andrewrk]

weakvoid*dlopen(constchar*name, intflags)
{
error="dlopen not implemented";
returnNULL;
}
weakvoid*dlsym(void*library, constchar*name)
{
error="dlsym not implemented";
returnNULL;
}

oh.

ok, well, it won't be very difficult to implement this in zig 🙄

[alexrp]

It is just another one of those "emulated" WASI libraries that do nothing so existing software can compile. For whatever reason it doesn't follow the usual naming pattern.

FWIW, these all strike me as silly and I wouldn't be opposed to removing them, but that'd probably be separate work.

[jedisct1]

Most notable improvements include realpath and stubs for statvfs, chmod, etc.

Other changes are for WASI 0.2 which is going to be painful to support since 3rd party tools written in Rust may become necessary in addition to LLVM (WebAssembly/wasi-sdk#402)

[jedisct1]

This also removes junk we were shipping but irrelevant to webassembly.

[lin72h]

10KLoc reduce, feel so good

[jedisct1]

Added the wasi-libc libdl shim, so that tools such as py2wasm can use zig cc.

[andrewrk]

man. wasi-libc kinda sucks. how about we provide libc via zig code instead?

[jedisct1]

wasi-libc is horrible. And it's getting worse with the wasi preview 2 bits.

But rewriting it in Zig would be a lot of work.

[alexrp]

@jedisct1 I've rebased this on master locally, do you mind if I force-push to your branch? I'd like to get this in for 0.14.0.

[jedisct1]

Sure, go ahead!

There's been changes in wasi-libc since that PR, but the build system has become a giant spaghetti mess.

[alexrp]

Yeah, I figure this is likely to be our last wasi-libc update, what with all the Rust tooling they're apparently mixing in due to the component model stuff. #2879 is probably going to be the way forward in the long term (unless we develop our own component model tooling).

[alexrp]

@andrewrk you ok with taking this as our (probably) last wasi-libc update? (At least unless/until we develop self-hosted component model tooling.)

[jedisct1]

wasi-libc still supports wasi.0.1 (aka wasi-core, wasip1, etc). So it can still be updated; I don't think the Rust tools are required for it.

[andrewrk]

you ok with taking this as our (probably) last wasi-libc update?

Yes 👍

[andrewrk]

this strikes me as strange... wasi doesn't support dynamic linking, so what in the world is this library?

[andrewrk]

weakvoid*dlopen(constchar*name, intflags)
{
error="dlopen not implemented";
returnNULL;
}
weakvoid*dlsym(void*library, constchar*name)
{
error="dlsym not implemented";
returnNULL;
}

oh.

ok, well, it won't be very difficult to implement this in zig 🙄

[alexrp]

It is just another one of those "emulated" WASI libraries that do nothing so existing software can compile. For whatever reason it doesn't follow the usual naming pattern.

FWIW, these all strike me as silly and I wouldn't be opposed to removing them, but that'd probably be separate work.

[jedisct1]

Most notable improvements include realpath and stubs for statvfs, chmod, etc.

Other changes are for WASI 0.2 which is going to be painful to support since 3rd party tools written in Rust may become necessary in addition to LLVM (WebAssembly/wasi-sdk#402)

[jedisct1]

This also removes junk we were shipping but irrelevant to webassembly.

[lin72h]

10KLoc reduce, feel so good

[jedisct1]

Added the wasi-libc libdl shim, so that tools such as py2wasm can use zig cc.

[andrewrk]

man. wasi-libc kinda sucks. how about we provide libc via zig code instead?

[jedisct1]

wasi-libc is horrible. And it's getting worse with the wasi preview 2 bits.

But rewriting it in Zig would be a lot of work.

[alexrp]

@jedisct1 I've rebased this on master locally, do you mind if I force-push to your branch? I'd like to get this in for 0.14.0.

[jedisct1]

Sure, go ahead!

There's been changes in wasi-libc since that PR, but the build system has become a giant spaghetti mess.

[alexrp]

Yeah, I figure this is likely to be our last wasi-libc update, what with all the Rust tooling they're apparently mixing in due to the component model stuff. #2879 is probably going to be the way forward in the long term (unless we develop our own component model tooling).

[alexrp]

@andrewrk you ok with taking this as our (probably) last wasi-libc update? (At least unless/until we develop self-hosted component model tooling.)

[jedisct1]

wasi-libc still supports wasi.0.1 (aka wasi-core, wasip1, etc). So it can still be updated; I don't think the Rust tools are required for it.

[andrewrk]

you ok with taking this as our (probably) last wasi-libc update?

Yes 👍

[andrewrk]

this strikes me as strange... wasi doesn't support dynamic linking, so what in the world is this library?

[andrewrk]

weakvoid*dlopen(constchar*name, intflags)
{
error="dlopen not implemented";
returnNULL;
}
weakvoid*dlsym(void*library, constchar*name)
{
error="dlsym not implemented";
returnNULL;
}

oh.

ok, well, it won't be very difficult to implement this in zig 🙄

[alexrp]

It is just another one of those "emulated" WASI libraries that do nothing so existing software can compile. For whatever reason it doesn't follow the usual naming pattern.

FWIW, these all strike me as silly and I wouldn't be opposed to removing them, but that'd probably be separate work.

[jedisct1]

Most notable improvements include realpath and stubs for statvfs, chmod, etc.

Other changes are for WASI 0.2 which is going to be painful to support since 3rd party tools written in Rust may become necessary in addition to LLVM (WebAssembly/wasi-sdk#402)

[jedisct1]

This also removes junk we were shipping but irrelevant to webassembly.

[lin72h]

10KLoc reduce, feel so good

[jedisct1]

Added the wasi-libc libdl shim, so that tools such as py2wasm can use zig cc.

[andrewrk]

man. wasi-libc kinda sucks. how about we provide libc via zig code instead?

[jedisct1]

wasi-libc is horrible. And it's getting worse with the wasi preview 2 bits.

But rewriting it in Zig would be a lot of work.

[alexrp]

@jedisct1 I've rebased this on master locally, do you mind if I force-push to your branch? I'd like to get this in for 0.14.0.

[jedisct1]

Sure, go ahead!

There's been changes in wasi-libc since that PR, but the build system has become a giant spaghetti mess.

[alexrp]

Yeah, I figure this is likely to be our last wasi-libc update, what with all the Rust tooling they're apparently mixing in due to the component model stuff. #2879 is probably going to be the way forward in the long term (unless we develop our own component model tooling).

[alexrp]

@andrewrk you ok with taking this as our (probably) last wasi-libc update? (At least unless/until we develop self-hosted component model tooling.)

[jedisct1]

wasi-libc still supports wasi.0.1 (aka wasi-core, wasip1, etc). So it can still be updated; I don't think the Rust tools are required for it.

[andrewrk]

you ok with taking this as our (probably) last wasi-libc update?

Yes 👍

[andrewrk]

this strikes me as strange... wasi doesn't support dynamic linking, so what in the world is this library?

[andrewrk]

weakvoid*dlopen(constchar*name, intflags)
{
error="dlopen not implemented";
returnNULL;
}
weakvoid*dlsym(void*library, constchar*name)
{
error="dlsym not implemented";
returnNULL;
}

oh.

ok, well, it won't be very difficult to implement this in zig 🙄

[alexrp]

It is just another one of those "emulated" WASI libraries that do nothing so existing software can compile. For whatever reason it doesn't follow the usual naming pattern.

FWIW, these all strike me as silly and I wouldn't be opposed to removing them, but that'd probably be separate work.