| apw | awreece, it must be visible somewhere check-point restore can get it out | 10:17 |
|---|---|---|
| TJ- | apw: finally managed to a get a kernel build running for bisect; had to fight out-of-date/incomplete instructions in the Wiki :) Once I'm sure it's all completing correctly I'll bring the wiki up-to-date | 10:19 |
| apw | TJ-, great, those of us who know how no longer see the errors | 10:20 |
| TJ- | indeed... it's a few years since I focused on Ubuntu kernel tools :) | 10:21 |
| TJ- | spoke too soon! "ld: final link failed: No space left on device" | 10:23 |
| apw | now that i don't think is our fault :) | 10:23 |
| * TJ- sniggers | 10:24 | |
| TJ- | apw: Oh it is - you should make sure it doesn't gobble up 10GB during a single-arch single-flavour build :) | 10:25 |
| TJ- | apw: thankfully we have lvextend and resize2fs | 10:25 |
| apw | TJ-, life savers for sure | 10:26 |
| TJ- | apw: before I start digging, any ideas on this failure? "ld: BFD (GNU Binutils for Ubuntu) 2.24 internal error, aborting at ../../bfd/merge.c line 877 in _bfd_merged_section_offset" | 10:40 |
| apw | TJ-, other than its the compiler barfing on its own input, no | 10:57 |
| apw | not seen that specifici fail | 10:57 |
| apw | are you building in a clean chroot for the series this is being build for ? | 10:57 |
| TJ- | apw: I think maybe that was a by-product of the out-of-disk space... probably a truncated object file. I just restarted the build from clean. I'm not building in a chroot | 10:59 |
| apw | TJ-, yeah could easily be indeed, i always use a chroot, they are so much more predictable | 11:56 |
| TJ- | Had another bisect build fail: "II: Checking modules for lowlatency...previous or current modules file missing!" despite starting the build with AUTOBUILD=1 which, in "debian/rules.d/0-common-vars.mk" is supposed to set skipmodule, skipabi, etc., to true *but* the fail is because $(skipmodule) isn't set in "debian/scripts/module-check" (line 21) when it is called from debian/rules.d/4-checks.mk (line 16) - is something else going on I'm not aware of ? | 12:04 |
| TJ- | Despite the AUTOBUILD=1 I've had to add "skipabi=true skipmodules=true" to get those options recognised | 12:11 |
| TJ- | Grrr! I'm going back to bed! right at the end, packaging: "error in 'Version' field string '3.16.0-1.1~bisect_01': invalid character in revision number" ... like playing snakes and ladders :s | 12:14 |
| apw | TJ-, the autobuild=1 has likely been broken as its not somehting we commonly use .. | 12:19 |
| apw | TJ-, yeah some of those things it could check earlier, we did add some early checks for config | 12:19 |
| TJ- | apw: I'll add them as mini projects to me TODO list since all the kernel build documentation conspicuously refers to AUTOBUILD | 12:29 |
| apw | TJ-, feel free to file a bug against linux saying it AUTOBUILD doesn't seem to work, and let us know the bug # here | 12:29 |
| TJ- | apw: I'll do that once I've figured out the fix and provide a patch | 12:37 |
| apw | TJ-, at | 12:45 |
| apw | ta even | 12:45 |
| TJ- | Looks like Steve Conklin's advice in KernelBisection to add a ~xyz to the version number breaks package creation. Getting errors because the full version string isn't being found in 'control' - because the version assigned in 'rules' $(abinum) and used to replace ABINUM in the templates only contains <kernel-version>-<ABI>.<upload> and drops any ~xyz but the build system does use the ~xyz in creating directories, version strings in object code, etc. | 14:43 |
| TJ- | I'll fix up those docs; the short answer is always use "fakeroot debian/rules AUTOBUILD=1 ..." *not* "AUTOBUILD=1 fakeroot debian/rules ..." | 14:53 |
| TJ- | Next bug... building "binary-indep" (to generate the independent header package) fails due to 'do_tools_usbip=true' (set from 'debian.master/rules.d/amd64.mk') but supporting file "$(builddirpa)/tools/usb/usbip/autogen.sh" is missing. Had to add "do_tools_usbip=false" to the 'debian/rules' command-line | 15:58 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!