[Jun 22, 20:39]
<jordan> mikedld, ping
<jordan> I'm looking at 4160-01-error.patch
<jordan> and the changes to tr_strdup_vprintf()
<jordan> the va_copy is a nice improvement, I hadn't thought about the possible safety issues of making multiple calls to evutil_vsnprintf() with the same va_list
<jordan> I wonder if it might be better (simpler, terser, "cleaner") to implement it in terms of evbuffer_add_vprintf() and get rid of that 2 KiB statbuf
<jordan> eg, something like
<jordan> char *
<jordan> tr_strdup_vprintf (const char * fmt, va_list ap)
<jordan> {
<jordan> struct evbuffer * buf = evbuffer_new ();
<jordan> evbuffer_add_vprintf (buf, fmt, ap);
<jordan> return evbuffer_free_to_str (buf);
<jordan> }
* zz_sidv is now known as sidv
<jordan> mikedld, (2) in tr_error_free()'s #ifdef TR_PARANOID clause, maybe setting code and message to a DEADBEEF style value rather than zeroing it all out
<jordan> mikedld, (3) this is pretty minor, but prefer tr_new() to tr_malloc()
<jordan> mikedld, (4) this is debatable, but IMO there's no need to check for new/malloc failure... if we get to the point where pulling a struct off the heap fails, then nothing else is going to work anyway
<jordan> there's a case to be made for failing gracefully but if libtransmission can't make heap allocations then it's not going to be able to upload or download either
<jordan> so failing gracefully would mean "keep running but never do anything" ...
<jordan> in that hypothetical state, I'd argue it's better to crash and let the crash reporter tool file a bug so that we could do a postmortem and see what went wrong