On Fri, Sep 02, 2022 at 05:20:09PM +0200, Omar Polo wrote:
> commit #2 and #4 are real leaks, #3 only on the error path.  #1 is
> just a redundant check I guess.

ok

> however, the blame page still leaks like crazy.  There's some serious
> leak between got_gotweb_blame_cb and got_output_file_blame that I'm
> just not seeing.
> 
> In a controlled environment (localhost) after 1K requests to a blame
> page (always the same to rule out the cache) gotwebd' RES is 24M.  If
> I comment the call to got_blame then it's "just" ~5M.  So, it leaks
> both outside got_output_file_blame but the more serious leak is
> inside.

Is this easy to trigger via a reproducer hacked into got/got.c?