Summary: | DAHLIN-00337: Unable to build DAHDI-Linux 2.9.1 in mock chroot | ||
Reporter: | Anthony Messina (amessina) | Labels: | |
Date Opened: | 2014-03-28 19:56:47 | Date Closed: | 2014-03-31 16:31:05 |
Priority: | Major | Regression? | |
Status: | Closed/Complete | Components: | General |
Versions: | Frequency of Occurrence | Constant | |
Related Issues: | |||
Environment: | Both Fedora 20 and CentOS 6.5 Koji/mock chroots | Attachments: | ( 0) dahdi-linux-2.9.1-hotplug-install-fix.diff |
Description: | Unfortunately, after
http://git.asterisk.org/gitweb/?p=dahdi/linux.git;a=commitdiff;h=6cebc1c1fb12cc0661f3810ef47ad33206b2e398 I am unable to build DAHDI-Linux in a mock chroot for packaging purposes. I believe this is related to the Makefile calling install_firmware with only 2 args, where install_firmware is a shell script with DESTDIR set to $3, which is empty. In this case, the DESTDIR evaluates to "/usr/lib/hotplug/firmware", rather than "<buildroot_destdir>/usr/lib/hotplug/firmware". make -C drivers/dahdi/firmware hotplug-install DESTDIR=/builddir/build/BUILDROOT/dahdi-linux-2.9.1-1.fc20.x86_64 HOTPLUG_FIRMWARE=yes make[1]: Entering directory `/builddir/build/BUILD/dahdi- linux-2.9.1/drivers/dahdi/firmware' mkdir -p /builddir/build/BUILDROOT/dahdi- linux-2.9.1-1.fc20.x86_64/usr/lib/hotplug/firmware mkdir -p /builddir/build/BUILDROOT/dahdi- linux-2.9.1-1.fc20.x86_64/lib/firmware Installing dahdi-fw-oct6114-032.bin to hotplug firmware directories install: cannot create regular file '/usr/lib/hotplug/firmware': No such file or directory make[1]: *** [hotplug-install] Error 1 make[1]: Leaving directory `/builddir/build/BUILD/dahdi- linux-2.9.1/drivers/dahdi/firmware' make: *** [install-firmware] Error 2 | ||
Comments: | By: Elwin Andriol (e.andriol) 2014-03-29 13:56:00.032-0500 I was able to confirmed this issue by building a .deb package, with pbuilder on Debian Wheezy. However, instead of [6cebc1c|http://git.asterisk.org/gitweb/?p=dahdi/linux.git;a=commitdiff;h=6cebc1c1fb12cc0661f3810ef47ad33206b2e398], it is git commit [7ce8498|http://git.asterisk.org/gitweb/?p=dahdi/linux.git;a=commitdiff;h=7ce849846502ff9d5a6a7a95cf89a7f23c5680ef] that is the actual cause of the problem. The issue is not specific to any OS or building environment. As the OP mentioned, the install_firmware scripts expects the DESTDIR variable (essentially you installation prefix) as its third parameter, but the Makefile only supplies 2 parameters on each invocation. As long as things are being installed in the system's root directory, this issue does not show because DESTDIR is an empty variable in that case. It is nonetheless a serious bug that will cause trouble for anybody involved in packaging DAHDI-linux. By: Elwin Andriol (e.andriol) 2014-03-29 14:16:13.715-0500 *dahdi-linux-2.9.1-hotplug-install-fix.diff* should fix the problem By: Tzafrir Cohen (tzafrir) 2014-03-30 11:02:32.171-0500 The issue seems to be 7ce849846502ff9d5a6a7a95cf89a7f23c5680ef . Add an extra DESTDIR parameter for each? By: Shaun Ruffell (sruffell) 2014-03-31 09:06:48.294-0500 Uggh..my bad. I was out last week, but I see that Tzafrir has sent a patch. I'll see about getting this rolled out. By: Elwin Andriol (e.andriol) 2014-03-31 09:54:01.050-0500 I assume the bug fix will be in the git repo soon, but that will only solve the problem for people who build straight from git master. For many (automatic) package builders who use the latest release version (currently 2.9.1) this bug will be a show stopper, until the next release. That is, unless they integrate a special hotfix patch of their own of course (like I already did). Considering this consequence, would it be suggestible to bump up the current release version with a hotfix included? I do understand that there may be many reasons for being conservative about such a move. I'm just asking. For me the problem is already fixed, so the suggestion is more for the sake of preventing other people from getting annoyed by no longer being able to build DAHDI-Linux packages. By: Shaun Ruffell (sruffell) 2014-03-31 10:48:25.405-0500 Yes, I agree with you. I think we'll see a 2.9.1.1 here shortly. I just need to coordinate with rmeyerriecks this morning. By: Russ Meyerriecks (rmeyerriecks) 2014-03-31 14:21:58.608-0500 v2.9.1.1 tagged with the fix for this http://git.asterisk.org/gitweb/?p=dahdi/linux.git;a=commitdiff;h=22f034a307bf0ccdf63bba6aa0f72f8dbcba69c9 release tarballs to follow shortly |