On Mon, Jun 29, 2020 at 11:16:54AM +0200, Graham Leggett wrote: > On 29 Jun 2020, at 09:19, Joe Orton <jor...@redhat.com> wrote: > > > The litmus tests are failing, not the perl-framework tests: > > > > https://travis-ci.org/github/apache/httpd/jobs/702768269#L2491 > > Ah - that’s where I was going wrong. > > > You can also see that there are segfaults logged from line 2519 onwards: > > > > https://travis-ci.org/github/apache/httpd/jobs/702768269#L2519 > > > > Running litmus locally, the backtrace looks like this: > > So the simple workaround is to be defensive against ctx->r being NULL, > but I need to check - is there ever a legitimate reason for there to > be a filled out ctx structure including a filename, but with no > request?
I don't know, you added the ctx->r field in r823703 and it's not obvious exactly what the semantics are supposed to be. If the request_rec is supposed to correspond to the dav_resource object then it makes sense that it is NULL when looking at the dav_resource representing the parent collection (in this case) or some other resource during a walk. It would be good to document this! Regards, Joe