Hi, * Erik Hovland wrote:
> >> - FREE (tcode); /* __FREE_CHECKED__ */ > >> + FREE (&tcode); /* __FREE_CHECKED__ */ > > > > The same for the use of FREE(). This is probably wrong. > I'll have to look again. But if I recall tcode is allocated and since > FREE takes the pointer to the pointer, the & is required. But I am not > in front of my workstation right now. So I will defer to your > guidance. My fault, I didn't check it thoroughly enough. tocode is allocated as an array and should be FREE(&tocode). Valgrind also reports a leak on fcode so I fixed that, too plus the other leaks as b5cbd0dab863. I've separated the fpin change to give bisect a chance in case something goes wrong, it's pushed as 4fe35e9984ba. Thanks. > >> - CONTEXT f; > >> + CONTEXT *f = NULL; > > > > Here I don't understand what the leak should be. It just changes from > > stack to allocated memory causing more overhead I think. Can you explain > > why you change this? > Heh, the problem here is that the function that clears CONTEXT f > returns a pointer as an error indicator. [...] I see. mx_open_mailbox() allocates memory if you pass a null pointer. Shouldn't that also be released somewhere? Rocco
