Specific changes include 1. Using shared variables and locks where appropriate for test pass/fail, etc. state. 2. Making output from running tests go through a central function so that it can be buffered for each thread and output in a sensible order. 3. Rearranging some code and variables to group together things that would be thread specific from things that are global. 4. Change countFinishedTest to work better in an environment when the total number of tests to run isn't known. Since each test thread will pick off as many tests as it can, it isn't known ahead of time how many tests will run per thread.
Created attachment 29386 [details] Proposed fix. WebKitTools/ChangeLog | 19 ++++ WebKitTools/Scripts/run-webkit-tests | 194 ++++++++++++++++++++-------------- 2 files changed, 132 insertions(+), 81 deletions(-)
Comment on attachment 29386 [details] Proposed fix. On hold for now.
Removed myself from this bug as I will have no time for the foreseeable future to do anything on this. Eric may be starting to do something though.
The current perl script isn't going to be refactored by me. I suspect it will be replaced by the python version which is underway.