Crash_Override/OVMS3-idf
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

newlib locking: Turns out the "hack" is the way to make stdout thread-safe in newlib

Browse source
This commit is contained in:
Angus Gratton 2016-08-24 18:09:51 +08:00
parent 93c92f7a5b
commit 4b281af0f7
1 changed files with 13 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

18
components/esp32/syscalls.c
View file

@ -138,12 +138,20 @@ int _open_r(struct _reent *r, const char * path, int flags, int mode) {
ssize_t _write_r(struct _reent *r, int fd, const void * data, size_t size) { ssize_t _write_r(struct _reent *r, int fd, const void * data, size_t size) {
const char* p = (const char*) data; const char* p = (const char*) data;
if (fd == STDOUT_FILENO) { if (fd == STDOUT_FILENO) {
/* THIS IS A HACK!!! The stdout "file" should be locked while static _lock_t stdout_lock; /* lazily initialised */
this code is called (it's locked fflush.c:280 before /* Even though newlib does stream locking on stdout, we need
__sflush_r is called.) It shouldn't be necessary to a dedicated stdout UART lock...
re-lock, but due to some unknown bug it is...
This is because each task has its own _reent structure with
unique FILEs for stdin/stdout/stderr, so these are
per-thread (lazily initialised by __sinit the first time a
stdio function is used, see findfp.c:235.
It seems like overkill to allocate a FILE-per-task and lock
a thread-local stream, but I see no easy way to fix this
(pre-__sinit_, tasks have "fake" FILEs ie __sf_fake_stdout
which aren't fully valid.)
*/ */
static _lock_t stdout_lock;
_lock_acquire_recursive(&stdout_lock); _lock_acquire_recursive(&stdout_lock);
while(size--) { while(size--) {
#if CONFIG_NEWLIB_STDOUT_ADDCR #if CONFIG_NEWLIB_STDOUT_ADDCR