| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When performing an out-of-source build (in particular, with "make
distcheck"), the rules that we were using to replace @LIBDIR@ and
friends were failing. Automake does a lot of path magic in its rules,
but not within the shell commands themselves, so ultimately we wound
up with some confusion between the source and build directories.
Rather than hack around this problem in the top-level Makefile.am,
I've converted the project to a recursive build, where the top-level
Makefile.am delegates to Makefile.am in the subdirectories. This
over-complicates things, but does fix the out-of-source build. Why?
A guess: because the path magic works better in "." than elsewhere.
Signed-off-by: Michael Orlitzky <mjo@gentoo.org>
|
|
|
|
|
|
| |
The $(srcdir) in Makefile.am is implicit. Let's keep it that way.
Signed-off-by: Michael Orlitzky <mjo@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
| |
Since all of the log files for php-fpm wind up in the same place
regardless of slot, we install a single logrotate file as part of
eselect-php. The contents of the file may not be perfect -- they were
taken from the example submitted on bug 692194 untested -- but will be
easy to update in the future.
Bug: https://bugs.gentoo.org/692194
Signed-off-by: Michael Orlitzky <mjo@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
The tmpfiles.d entry for php-fpm creates one "runstate" directory for
all slots of php-fpm. Since none of that depends on the currently
selected slot, we ship it as part of eselect-php rather than
dev-lang/php which would either install one copy for each slot.
Signed-off-by: Michael Orlitzky <mjo@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
While OpenRC services are powerful enough to compute the selected
version of php-fpm, the corresponding systemd services are not and
must use a fixed path. The executable at that fixed path must
therefore decide which version of php-fpm to run before running it.
This commit adds a script to do just that.
This script is already present in the app-eselect/eselect-php
package, tacked on from $FILESDIR. It's a better fit as part of
the repository.
Signed-off-by: Michael Orlitzky <mjo@gentoo.org>
|
|
|
|
|
|
|
| |
We're already substituting a few other GNU directories, and we'll
need the libexecdir in the future.
Signed-off-by: Michael Orlitzky <mjo@gentoo.org>
|
|
|
|
|
|
|
| |
OpenRC's init and conf files need to be named the same, and so we hit
problems trying to store them both in the "openrc" directory. This
commit puts the conf file in a "conf.d" subdirectory and the init
script in an "init.d" subdirectory.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
The OpenRC init script is now installed by the Makefile, so to be
useful it needs to be called "php-fpm". To that end, I've renamed both
the init script and the config file to lack the word
"example". Moreover, now that these files are code-like and not
documentation-like, I've renamed the directory containing them from
"doc" to "openrc".
|
| |
|
|
|
|
|
|
|
|
| |
We cannot rely on the assumption that /usr/lib will always point
to the right location, Bug 624528 demonstrates this.
There is talk in Bug 506276 of making /usr/lib a generic target
for all arches and stop supporting the symlink.
|
| |
|
|
|
|
|
| |
Without the extra hint, automake was omitting 70_mod_php.conf.in from
the tarball.
|
|
|
|
|
|
|
|
| |
The latest eselect-php ebuild allows users to avoid 70_mod_php.conf by
setting USE="-apache2". Now that the conf file is part of eselect-php,
we replicate that ability with --enable-apache2=no, or in fact by
default. To install the apache conf file, one passes --enable-apache2
to the configure script.
|
|
|
|
|
|
|
|
|
|
| |
The last few commits introduced a problem: by passing
php.eselect.in.in through autoconf, we unwittingly replace things like
@bindir@ with ${exec_prefix}. This is because lowercase @bindir@
gets defined for automake and is intended to be based on the other
directory variables like @exec_prefix@. Since we're replacing them
after the fact in the Makefile, this commit avoids the autoconf
mangling by making them all uppercase.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We are currently shipping 70_mod_php.conf as part of the Gentoo ebuild
for eselect-php. This introduces a potential disconnect between the
path in php.eselect and the path in 70_mod_php.conf. By adding the
conf file to the project, we are able to ensure that they are both
set to the same value. Moreover, the new autotools directory magic
lets us set them both based on @localstatedir@. So, for example, in
the conf file we have "@localstatedir@/lib/eselect-php/mod_php.conf"
and in php.eselect we have "@localstatedir@/lib/eselect-php".
We have followed the PHP project's lead in assuming that
$localstatedir will be set to (for example) /var and not /var/lib.
See Gentoo bug 572002. For testing, you should now use something like,
./configure --bindir=/usr/bin --sysconfdir=/etc --localstatedir=/var
Of course, the ebuild for eselect-php will automatically pass the
correct values to ./configure.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When we create php.eselect from php.eselect.in using autoconf, we
don't have a way to get the full, expanded value of @bindir@,
@libdir@, and the other paths that we need. The autoconf documentation
suggests a workaround for this: add another layer of indirection, and
use the Makefile to substitute those values into the output files.
This commit sets the stage by renaming php.eselect.in to
php.eselect.in.in. The first round of processing by autoconf takes
php.eselect.in.in to php.eselect.in. The Makefile will then generate
php.eselect from php.eselect.in after substituting @bindir@, @libdir@,
and @localstatedir@.
|
|
|