MD5sum usage in tests means I can't compare old to new

[gitoliver]

Here is the output for one of the tests:

Running the following command for test MinGasPdb: 
    md5sum /website/TESTS/git-ignore-me/pre-push/sequence/cb/Builds/000dade0-1fdd-470c-a272-cf573e48f2df/New_Builds/e6c2e2e8-758b-58b8-b5ff-d138da38dd22/min-gas.pdb | cut -d ' ' -f1
the answer is : 
>>>c37c6bce493eabeffe41bafdb8771cb0<<<
the CORRECT answer is : 
>>>050c7547d47239ace3dad7d34735ab0b<<<

I often change thing in gmml that means something about the output is different. I'll often have context to know that e.g. index numbering changes are ok, but if the 3D structure isn't exactly the same something is up and I need to investigate. At the gmml level we store the old output and do a diff. This makes is super easy to review diff outputs or load both structures into VMD and update the correct output. With just an md5sum I have no idea. I get that it reduces space, but I would reserve it for cases where the files are large enough to necessitate the inconvenience. Right now I just eyeball the new one and say "sure", which is going to bite us at some point.

[Lachele]

Although the test compares only the MD5 sum, I have been saving and committing the entire testing tree for 008. But, it appears that it was deleted from the repo.

[Lachele]

It is good to save the entire 'cb' tree because the test is pretty thorough.

[gitoliver]

I've been updating the md5sum as I go, I likely didn't sync the entire directly output at some point, realized they were out of sync and just deleted it as it's less than useful at that point.
I propose we go back to saving the entire output and compare directly to files within it without the md5sum step.

[Lachele]

The test compares entire directories, etc., so it's not just a couple files. It's entire states at each step in the tests. The md5sum keeps the test from being crazy to read. We can take the md5sum of the file in the saved directory (on the fly) bc that will help direct the tester to the location of the correct file.

…

On Mon, Dec 4, 2023 at 8:45 AM Oliver ***@***.***> wrote: I've been updating the md5sum as I go, I likely didn't sync the entire directly output at some point, realized they were out of sync and just deleted it as it's less than useful at that point. I propose we go back to saving the entire output and compare directly to files within it without the md5sum step. — Reply to this email directly, view it on GitHub <#78 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACCV7D2ASSUPBRVP6ZL4YJDYHXHYXAVCNFSM6AAAAABAF6XMXGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMZYGY3TIMJXGI> . You are receiving this because you were assigned.Message ID: ***@***.***>

[gitoliver]

Lachele is this implemented for specific files?

[Lachele]

I don't really understand what you mean by "for specific files". Some tests span multiple files, for example, this test:

008.0.testing-arrays.bash: [ListRSeqsSeqID]="/bin/ls -R ${sequenceSequencesPath}/${theCorrectSequenceID}"

ensures that all the expected files at least exist.

I made it easy to update the reference files. This capability is only in the gems-test branch because I've been being terrible about keeping the branch tags/rebases up to date. I'll work on this next deployment. I also plan to switch to a more GitHub-friendly workflow one day.

https://github.com/GLYCAM-Web/gems/blob/gems-test/tests/correct_outputs/008_README.md

It helps to have all the files because if something goes wrong, it's nice to see exactly how everything is supposed to be.