On 1/19/06, Norris Leong <[EMAIL PROTECTED]> wrote: > Hi All, > > I suspect that read_with_timeout (in file_io/win32/readwrite.c) may drop > data if > > (i) WaitForSingleObject (line 82) returns WAIT_TIMEOUT, and > (ii) Data arrived before the execution arrives at CancelIo (line 105). > > Tried the attached patch and the problem seems solved. I'm a novice of > Windows' file mgmt API and would appreciate if somebody can validate > this patch.
This seems reasonable to me, although I'm not anywhere close to a win32 expert, so I can't say for sure. Have you actually seen this race in practice, or is it just a theoretical possibility? I'm only asking because I'd be much more comfortable applying a patch if the contributor can say "this actually fixes a reproducable problem" ;-) -garrett
