On Fri, Oct 14, 2016 at 11:16 AM, Eric Covener <cove...@gmail.com> wrote:
> This was not backported and popped up in PR60251.
> Bill, can you have a look including my guess that it really should
> just be "temp_sa = r->useragent_addr;"?
While that code should *not* be triggered before r->useragent_addr
has been populated, some off-beat perl code causes these phases
to run out-of-sequence and we segfault not long after if this is run
without a post read request hook.
I blame a bad mod_perl example, but the cycle wasted to confirm
that useragent_addr is non-null isn't worth trimming.