GNOME Bugzilla – Bug 776277
TypeError: could not get a reference to type class in generate-setting-docs.py
Last modified: 2017-01-25 17:58:20 UTC
I'm not sure what changed here, since I tested this just a couple days ago and it seemed to be working fine, but NetworkManager is broken in JHBuild again. The error is going to be pasted in the next comment due to bug #776275. It happens with both python2 and python3.
/home/mcatanzaro/src/jhbuild/checkout/NetworkManager/libnm/generate-setting-docs.py:191: Warning: cannot retrieve class for invalid (unclassed) type 'void' setting = new_func() Traceback (most recent call last):
+ Trace 237003
setting = new_func()
Hi, Can you share a log to the build? What are the full configure options to reproduce? What's the git commit ID of the failing build? Could be related to bug 776178
Created attachment 342221 [details] build log Build log (including all configure options set by JHBuild). Commit is d136f0fedc37b23c39955da7442cfdcd0db75114 "libnm/docs: clarify "ipvx.dns-priority" setting in documentation"
I'll drop that unrecognized --disable-wimax flag from JHBuild now.
I cannot reproduce this build failure :(
I can reproduce it always on a clean build, but I don't understand the python nor the gobject-introspection involved. :(
I am unable to reproduce this with Ubuntu16.10, following the jhbuild instructions from https://wiki.gnome.org/Newcomers/BuildGnome (and ensuring to actually build upstream master of NetworkManager).
Thomas, thanks for testing this out in JHBuild. I spun up a new Ubuntu VM today and found that, indeed, the build works fine. Really sorry for sending you on this wild goose chase with an Ubuntu VM, I'm sure that took a couple hours. :( Then I tried to test with JHBuild on Fedora. I've made sure I'm using a completely clean build directory (rm -rf ~/.cache/jhbuild/build/NetworkManager) and source directory (git clean -xfd). I can reproduce it reliably here. So for whatever reason, the bug occurs on Fedora and not Ubuntu. Next, I cleared the source directory with git clean -xfd again and tried a build without using JHBuild at all (on Fedora), starting with './autogen.sh --disable-json-validation'. This worked fine. Next, I tried a build outside JHBuild with configure options similar to those set by JHBuild: ./autogen.sh PYTHON=python3 --disable-json-validation --disable-more-warnings --with-crypto=gnutls And that went fine too. So the issue is somehow specific to JHBuild on Fedora. (Which is unfortunately what I use for making GNOME releases....) (In reply to Michael Catanzaro from comment #0) > I'm not sure what changed here, since I tested this just a couple days ago > and it seemed to be working fine, but NetworkManager is broken in JHBuild > again. I think what changed was I had been working in an Ubuntu VM to make sure it was fine there, then switched back to Fedora and found it was broken, and forgot I had switched operating systems.
I am also unable to reproduce on F25 (both my machine and a newly installed VM). Maybe you need to run `jhbuild sysdeps --install`? Well... actually, yesterday when testing with Ubuntu1610 I saw the failure once, but couldn't get a grip on it. Afterwards I was unable to reproduce it even by purging the entire home directory.
(In reply to Thomas Haller from comment #9) > I am also unable to reproduce on F25 (both my machine and a newly installed > VM). :/ > Maybe you need to run `jhbuild sysdeps --install`? 'jhbuild build NetworkManager' will fail if you are missing sysdeps required by NetworkManager or its dependencies. But sometimes a module is missing a dependency. At any rate, I have all sysdeps installed for all of GNOME. > Well... actually, yesterday when testing with Ubuntu1610 I saw the failure > once, but couldn't get a grip on it. Afterwards I was unable to reproduce it > even by purging the entire home directory. Shame....
I ran into this on F25 as well http://pastebin.com/BFgxueJs
Created attachment 343740 [details] Patch Seems like a gobject-introspection or glib bug to me. Here's a patch that works around it.
(In reply to Lubomir Rintel from comment #12) > Created attachment 343740 [details] > Patch > > Seems like a gobject-introspection or glib bug to me. > > Here's a patch that works around it. Could this be committed?
yes. master: https://cgit.freedesktop.org/NetworkManager/NetworkManager/commit/?id=6c96aafaa9a61ca4fe713551a4cd4b15fb7e8a63 https://cgit.freedesktop.org/NetworkManager/NetworkManager/commit/?id=22722b77322fb3ddc1a99509340e8175acddb154 and backported to all stable branches where it matters.