Pleat slip test

[broucaries]

Hi,

Tests take too many time and trip some time limit on buildd

Could you split test suite in small tests in order to get regular display on the console and avoid to kill the build task

Bastien

[magick]

Validating the image formats consume the most amount of time. There are 600+ format tests in memory then on disk for a pixel cache in memory, memory-mapped, or on disk (3600+ tests in all). These tests are necessary for proper validation testing. There are currently two TAP tests to validate in memory and a second to validate on disk. We can easily split these into 3 tests reducing the individual run-time by 1/3. However, the 75 TAP tests will still take the same elapsed time. We're not sure it makes sense to split the 600 format tests into individual TAP tests so each run in a fraction of a second. The only alternative we can think of is to have two different validation tests, one short (default), and one comprehensive. However, that does not sit well with us since validation tests are there to ensure the project is running properly and error free. Given ImageMagick has a huge number of features and supports over 100 image formats, a large number of regression tests are expected.

[broucaries]

Slip it in 3 tests it will be sufficient for now

Split the format test will also help distribution for instance hurd faill to pass test :
see https://buildd.debian.org/status/fetch. ... 1369944988

But we do not know the test that fail

[magick]

None of the tests should fail and many did on the hurd build. Take a look at tests/cli-pipe.log and tests/validate-formats-in-memory.log and see if there is a clue why its failing. We do not have access to hurd but as expected all the unit tests work for us under Redhat, CentOS, and Fedora.

[glennrp]

Validating the image formats consume the most amount of time. There are 600+ format tests in memory then on disk for a pixel cache in memory, memory-mapped, or on disk (3600+ tests in all). These tests are necessary for proper validation testing. There are currently two TAP tests to validate in memory and a second to validate on disk. We can easily split these into 3 tests reducing the individual run-time by 1/3. However, the 75 TAP tests will still take the same elapsed time. We're not sure it makes sense to split the 600 format tests into individual TAP tests so each run in a fraction of a second.

Splitting it into a small number of tests would help. With 3600+ tests, when I'm running on a remote machine I get disconnected during the long tests due to some sort of timeout. I now avoid that by replacing tests/validate.h with one that only does a few tests that interest me, but that isn't useful for the general distribution. Splitting them into 3 or 4 TAP tests would be fine for me.

[broucaries]

See http://www.harding.motd.ca/autossh/

I use it daily

Split in more than 3 tests will be better, but 3 is a good beginning

Bastien