Latest changes to bitbake has broken the communication with Toaster. [poky]$ git pull [poky]$ . oe-init-build-env [build]$ . toaster start noweb The system will start. Successful start. [build]$ bitbake core-image-minimal Gave up; Too many tries: timeout while attempting to communicate with bitbake server [build]$
It seems that I have recovered by (a) killing any running toaster, and (b) deleting the "_git___git.yoctoproject.org_poky_master" repo. The root cause appears to be a sync error between the versions of bitbake in the host Toaster and the cached repo directory, which makes sense give the massive recent changes.
The basic root cause is resolved with 11851. The previous comment about deleting clones turns out to solve the problem by the simple time delay, not any mismatched content. I did get this issue intermittently with non-Toaster usage, which makes it a bb problem. I believe that has or is being addresses with other bitbake-devel patches. *** This bug has been marked as a duplicate of bug 11851 ***
New information. If you create a new build directory, start Toaster, and then attempt a bitbake build, you will generally get a server timeout (or sometimes you will get EOFError an on File "/usr/lib/python3.4/multiprocessing/connection.py", line 389, in _recv). If instead you run the bitbake command (like "quilt-native") _first_ and then start Toaster, the bitbake command will generally start and run just fine. There is a wait or race condition with the bitbake sever when Toaster is the command line UI.
1) Moving the code around and adding lots of log messages, it appears that the timeout is whatever server call happens after the server call to start the build . 2) I can consistently get the timeout the first time or two (e.g. 'bitbake quilt-native'). When I do finally get a build to run, I can consistently re-trigger the timeout if I "rm -rf cache/* tmp/ sstate-control/" and restart a build. 3) I suspect that knotty does not have this problem since it does not send any sever commands after the build starts (the only exception is when there is a CTLR-C from the user).
I found this issue when I tried to start toaster in the commit 9ed748a542b520c1cb763d981969233c0f5efd4e, I solved it by killing any bitbake process and starting over.
I have filed this bug against bitbake since the issue appears to not be in the Toaster code: Bug 11898 - bitbake server timeout on first 'server.runCommand' call after build starts I am writing a simpler test case which I will base on knotty. I will try out David Lopez Barriba observation (thank you!) to see if that helps. In my tests I tried to remove all background python and bitbake tasks, and I also replicated with the "noweb" option to remove that variable. My results were all over the place. I will keep this bug open to track the issue and the bitbake progress.
The bug 11898 has been accepted and is being worked on by bitbake-devel. I am formally marking this as a duplicate. *** This bug has been marked as a duplicate of bug 11898 ***