Daniel Black says this issue was fixed in https://github.com/MariaDB/server/pull/251#discussion_r85670517

Travis part of the fix was merged too.

danblack svoj The challenge with Travis currently is that it times out way too often when 50 minute hard limit is hit. We do have ccache enabled, and the log confirms that the steps about ccache downloading and uploading works, but somehow during the build the cache does not seem to speed up anything.

Decreasing the parallel count at now in December based on a measuring done in September might not produce the same result anymore.

Also as the Debian packaging has been upstreamed form Debian the MYSQL_BUILD_* variables used in a related commit https://github.com/MariaDB/server/commit/e14bdcb81c88156ce5e413865f7fe124321120eb are not really effective. In the Travis matrix there are supposed to be now on GCC 5 and one GCC 6 build but in fact both run:

https://s3.amazonaws.com/archive.travis-ci.org/jobs/184726773/log.txt

-- The C compiler identification is GNU 4.8.4

-- The CXX compiler identification is GNU 4.8.4

https://s3.amazonaws.com/archive.travis-ci.org/jobs/184726774/log.txt

-- The C compiler identification is GNU 4.8.4

-- The CXX compiler identification is GNU 4.8.4

I recommend we don't merge any of these other Travis-CI change suggestions unless if is confirmed that they still actuallly work as intended and in general I suggest we don't make any Travis-CI changes other than those that consistently decrease the build time so that there would not be any random timeoouts in the future.

otto, you spotted very valid issues.

Since travis is broken anyway and there's nobody besides Daniel volunteer fixing it, I'd rather let him do it as he sees fit. I believe we'll get functional travis much faster than if we would just hold back patches for months.

otto In an effort to help understand the lack of impact from ccache I've added https://github.com/MariaDB/server/pull/281/commits/f6099e30eb2a75e0f4143c5c1fe0dc6c8ef7bee9 . I don't know if resetting the stats at the start is the right thing to do yet. We'll see what it looks like after a few runs though.

The effectiveness of MYSQL_BUILD_XX has been reintroduced with https://github.com/MariaDB/server/pull/281/commits/bb3179d7c228285962b056a85efc870c023ee98f. I hope this is acceptable. LAUNCHER was added as ccache links may not be there with non-distro packages.

https://github.com/MariaDB/server/pull/282/commits/83b881f5f3f71671c01e61e66748c99ce5e47fec separates the cache for gcc 5 and 6 to keep the concurrent update of the cache from invalidating each others build.

Quite right that the parallel count may not be right yet. We'll see how it goes and do a few samples to see if its currently.