Hi Simon, On Fri, 31 May 2024 at 16:07, Simon Steiner <simonsteiner1...@gmail.com> wrote: > The tests should pass on linux, I haven’t tested windows. I had a look. The failing tests are intermittent failures, to do with performance measurement. I think I remember reading that the resolution of the timer on Windows is low compared to Linux - anyway the tests calculate a score based on (task duration)/(reference duration) and sometimes one or other of these will be zero, with unfortunate consequences.On other occasions there is an unexpected performance improvement/regression because the duration flips from one value to another (e.g. 6 to 12 ms) due to the low resolution. I guess the code was written when machines were a bit slower and hence the durations were less likely to be small and affected by these issues.
I also found a problem in the tests run from a git checkout. The base 64 tests encode a test binary file to base64 but some of the test files were treated by git as text files and therefore had their line endings changed to CRLF when checked out on Windows. My fixes for these are in a branch https://github.com/pedro-w/xmlgraphics-batik/tree/windows-tests and I can make it into a PR if you are interested. I don't have linux but I tested on MacOS and all tests still passed. Thanks Pete --------------------------------------------------------------------- To unsubscribe, e-mail: batik-dev-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: batik-dev-h...@xmlgraphics.apache.org
