Summary: | version bump : dev-perl/libapreq : 1.2 => 1.3 / 2.02 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Stephane Loeuillet <leroutier> |
Component: | New packages | Assignee: | Gentoo Perl team <perl> |
Status: | RESOLVED FIXED | ||
Severity: | enhancement | CC: | apache-bugs, beu, djimenez, magnet, mcummings, perl, rendhalver, scy-bugs-gentoo, tkay |
Priority: | High | ||
Version: | 1.4 | ||
Hardware: | All | ||
OS: | Linux | ||
URL: | http://httpd.apache.org/apreq/ | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | 47875 | ||
Bug Blocks: | 35407, 37131, 59413, 61893 | ||
Attachments: |
libapreq2 ebuild
patch to allow setting httpd binary location Updated version of ebuild that works better. Patch to turn offf testing that seems incompatible with portage Patch to fix sandbox problem updated ebuild with most of the testing working libapreq-2.0-test1.patch libapreq-2.0-test2.patch |
Description
Stephane Loeuillet
2003-12-08 19:09:06 UTC
Created attachment 22452 [details]
libapreq2 ebuild
I have been fighting libapreq2 for three days straight now, and here's as far
as I've gotten. This package is a nightmare to get to work with Gentoo. This
version might work with sandbox turned off. You will need a patch to make
it understand that httpd is not always named that, coming below.
Created attachment 22453 [details, diff]
patch to allow setting httpd binary location
This patch is referenced by the ebuild above.
Created attachment 28766 [details]
Updated version of ebuild that works better.
This ebuild integrates in two additional patches and fixes a typo that kept the
Perl stuff from being built. It seems to fix the ebuild and make it work
cleanly. It fixes the sandbox problem. I had to turn off the testing in the
makefile because it did not like portage at all, and I didn't have time to fix
that.
There is a required perl module (ExtUtils::XSBuilder) that does not have a
portage entry, so currently you have to install that one with g-cpan.pl before
this will compile.
There also seems to still be a runtime bug with respect to Mason. I will submit
a patch if I find out it is related to this.
Created attachment 28767 [details, diff]
Patch to turn offf testing that seems incompatible with portage
This is one of the patches used by the ebuild I just submitted.
Created attachment 28768 [details, diff]
Patch to fix sandbox problem
This is also used by the new ebuild. It fixes the sandbox problem, so the thing
will compile and install cleanly under normal emerge (with ~x86 keyword, of
course)
The problem with Mason was not a problem with this. I just followed the Mason with Apache2 mini-HOWTO, and that fixed it. So, that seems to indicate that the ebuild is working with everything I have to test it against. I had to add a 'dodir /usr/lib/apache2/modules' before the make install so that 'cp .lib/mod_apreq.so ${D}/usr/lib/apache2/modules' didn't fail. I've also submitted an ebuild for ExtUtils-XSBuilder-0.25 here: http://bugs.gentoo.org/show_bug.cgi?id=47875 . web-apps folks - can you take a look at this ebuild since it involves web apps (ha) to be sure we're clean before going forward? I'll try and take a look sometime Monday. Best regards, Stu Created attachment 34843 [details] updated ebuild with most of the testing working I have been trying to get testing to work with portage. I think I've made some progress---all but one test (cgi.t) work, but this last one looks like it's beyond me. Since testing cannot be done as root, and ${WORKDIR} isn't writable by 'nobody', I copy everything to ${T}, and run apxs with -user portage -group portage. That seems to work just fine. The ebuild also references two new (small) patches that will follow. You might also need to ln -s mod_perl into apache2/modules. cgi.t still fails when make test is run from the ebuild. looking at the output of env/t/TEST -verbose, it seems that cgi-bin/test_cgi returns 0 instead of the count. portage does not report any access violations, but it's most certainly related to the sandbox, b/c when I run 'make test' manually, the test passes. The test also fails when 'make test' is run from sandboxshell. I 'strace -f -e file'd it, and I didn't find anything suspicious in terms of file access. However, comparing the output of a successful and an unsuccessful run, it looks as if when test_cgi is called from the sandbox, no environment variables are set (not even $QUERY_STRING). Here's the actual output: successful run: 4421 execve("/var/tmp/portage/libapreq-2.02_alpha02/temp/env/t/cgi-bin/test_cgi", ["/var/tmp/portage/libapreq-2.02_a"...], ["HTTP_HOST=localhost.radagast:852"..., "HTTP_USER_AGENT=libwww-perl/5.69"..., "PATH=/bin:/sbin:/usr/bin:/usr/sb"..., "SERVER_SIGNATURE=", "SERVER_SOFTWARE=Apache/2.0.49 (G"..., "SERVER_NAME=localhost.radagast", "SERVER_ADDR=127.0.0.1", "SERVER_PORT=8529", "REMOTE_ADDR=127.0.0.1", "DOCUMENT_ROOT=/var/tmp/portage/l"..., "SERVER_ADMIN=you@your.address", "SCRIPT_FILENAME=/var/tmp/portage"..., "REMOTE_PORT=42231", "GATEWAY_INTERFACE=CGI/1.1", "SERVER_PROTOCOL=HTTP/1.0", "REQUEST_METHOD=GET", "QUERY_STRING=a=ddddd;b=ddddd;c=d"..., "REQUEST_URI=/cgi-bin/test_cgi?a="..., "SCRIPT_NAME=/cgi-bin/test_cgi"]) = 0 unsuccessful run: 4334 execve("/var/tmp/portage/libapreq-2.02_alpha02/temp/env/t/cgi-bin/test_cgi", ["/var/tmp/portage/libapreq-2.02_a"...], [0]) = 0 I am out of ideas at this point, so if anyone has any suggestions, do let me know. Created attachment 34844 [details, diff]
libapreq-2.0-test1.patch
Created attachment 34845 [details, diff]
libapreq-2.0-test2.patch
This will not compile for me: *** result: OK /usr/sbin/apache2 -d /var/tmp/portage/libapreq-2.02_alpha02/temp/env/t -f /var/tmp/portage/libapreq-2.02_alpha02/temp/env/t/conf/httpd.conf -DAPACHE2 using Apache/2.0.50 (prefork MPM) waiting 60 seconds for server to start: .Syntax error on line 20 of /var/tmp/portage/libapreq-2.02_alpha02/temp/env/t/conf/extra.conf: Invalid command 'ScriptSock', perhaps mis-spelled or defined by a module not included in the server configuration !!! server has died with status 255 (t/logs/error_log wasn't created, start the server in the debug mode) make[1]: *** [run_tests] Error 143 make[1]: Leaving directory `/var/tmp/portage/libapreq-2.02_alpha02/temp/env' make: *** [env_test] Error 2 !!! ERROR: dev-perl/libapreq-2.02_alpha02 failed. !!! Function src_compile, Line 66, Exitcode 2 !!! (no error message) I would suggest adding a dependency towards ExtUtils-XSBuilder, which currently is not done. The emerge fails if no ExtUtils-XSBuilder is installed. well libapreq2 worked for me using maketest in FEATURES as well. So... we might be able to unmask it. Could be related or duplicated: http://bugs.gentoo.org/show_bug.cgi?id=81116 please see bug 77547 for an up-to-date ebuild libapreq-2.0.6 should be in the tree in the next little bit (read: within 12 hours max) Mass re-assign. Mass re-assign. Closing since we've long since bumped past 1.2/2.01 (and we aren't package.masking it anymore either) $ ls libapreq* libapreq: CVS Manifest libapreq-1.0-r3.ebuild libapreq-1.2.ebuild metadata.xml ChangeLog files libapreq-1.2-r1.ebuild libapreq-1.3.ebuild libapreq2: CVS Manifest libapreq2-2.04.03.ebuild libapreq2-2.07.ebuild ChangeLog files libapreq2-2.06.ebuild metadata.xml |