> On Feb 8, 2017, at 4:09 PM, Jacob Champion <champio...@gmail.com> wrote: > > On 02/08/2017 12:31 PM, Jim Jagielski wrote: >> Have you even TRIED it? > > Yes, with the latest trunk, a config of > > <FilesMatch "\.php$"> > SetHandler "proxy:fcgi://localhost:10102/" > </FilesMatch> > > leads to a SCRIPT_FILENAME of > 'proxy:fcgi://localhost:10102//tmp/apache-svn-trunk/htdocs/hello.php'. > >> I have. With PHP-FPM as well as that Perl script FPM. All works >> as expected. > > I understand that it works for the use cases you have tested. My change > worked for the use cases I tested, and Eric's follow-up change worked for the > use cases he tested, but we keep finding new regressions. This is a high-risk > code path, and the proposed backport makes it more complex instead of less > IMO. > > There's a good chance that the core of your addition will be part of the > eventual solution. My point is that without fixing the other stuff at the > same time, we've just added yet another type of almost-correct behavior that > we and the FCGI implementations have to code around. >
I am confused then... what else are you proposing? Eric's envar fix allows for people to basically adjust at their whim. What else is needed??