Use individual socket syscalls. #13272
Conversation
sbc100
force-pushed
the
remove_socketcall
branch
from
f03a4c9 to
07ff216
Compare
sbc100
force-pushed
the
remove_socketcall
branch
from
07ff216 to
0a3c359
Compare
Remove the SYSCALL_USE_SOCKETCALL definition which was causing musl to use a single syscall for all socket calls (a mostly legacy linux convention apparently). This is much better for emscripten because it allows DCE and more precise dependencies.
sbc100
force-pushed
the
remove_socketcall
branch
from
0a3c359 to
b54a861
Compare
|
@juj could you take a look? I hope you will be in favor of this as its win for DCE-ability. |
| #define __NR_sendmsg 370 | ||
| #define __NR_recvfrom 371 | ||
| #define __NR_recvmsg 372 | ||
| #define __NR_shutdown 373 |
There was a problem hiding this comment.
Does musl establish this numbering, or do we? Will this stay in sync with musl?
There was a problem hiding this comment.
Oh nm, reading emscripten-core/musl@c2feda4 suggests these are well established.
It looks like this was a recent change to musl upstream, did we update musl already before to support this? (or do we not really need to do so to update?)
There was a problem hiding this comment.
IIUC the recent change to upstream musl was to avoid socketcall multiplexing on more architecutres, but I think the idea was that any architectures that support the split out syscalls shoudl already have been using them. (i.e. it was arguably a bug that they were using socketcall).
Also, remove stray `err` logging from #13272.
Remove the SYSCALL_USE_SOCKETCALL definition which was causing
musl to use a single syscall for all socket calls (a mostly legacy
linux convention apparently).
For a little background on this see:
emscripten-core/musl@c2feda4
This is much better for emscripten because it allows DCE and more
precise dependencies.
The changes to library_syscall.js are just re-arranging code, there should be no
functions changes here.