On Fri, Apr 8, 2016 at 6:32 PM, William A Rowe Jr <wr...@rowe-clan.net> wrote: > It was working, in the sense that it had the intended effect (the [un]define > took effect) in the broader global context. > > This is a breaking change to some potentially existing configs, however > misguided they are, which is the sort of thing we've avoided in the released > branch. > > Could we log an error rather than preventing startup? One issue is that > these directives are encountered prior to opening the error log file. One > possible fix would be to have a second directive handler with the sole > purpose of emitting errors, running at the normal processing scope, and > not within exec_on_read.
This seems to work inside a directive handler on unix and ends up in the console: ap_log_perror(APLOG_MARK, APLOG_STARTUP, 0, cmd->pool > > This shouldn't be allowed to fester on trunk, obviously, but for 2.4 it > seems > like something we shouldn't alter, no matter how much it frustrates users > who used this unintentionally. That's fair, I don't remember if I intentionally or unintentionally didn't propose it immediately.