Problematic behavior with repeat submission of same patch (related to #218)

[atomrab]

I assume this issue is connected to #218, but this time with labels rather than dates (feel free to merge if preferable).

I made a local IDB and populated it with the Chronique des Fouilles dataset to change one label to match the authority vocabulary better.

Changed "Byzantin" to "Période byzantine", added new alternate label "Byzantin".

I submitted the patch, which registered both changes properly. Then I accidentally hit the "Reject button" (don't ask -- it was red and drew my eye).

I resubmitted the patch from the same local IDB, which shows "Période byzantine" as the new primary label; the change is properly noted, as below:

But when I review the new patch submission, it shows the added alternate label but NOT the changes to the primary label:

If I try to make a new IDB and change it again in the same browser, no changes at all register for submission. But if I try again in a different browser, I can see the changes when I try to submit -- so that part is probably a browser caching issue.

BUT: when I submit the patch through the second browser, with a clean history, the result is the same in the patch review -- the alternate label appears, but no changes to the primary label are registered. Since they registered the first time I submitted this patch (the accidentally rejected one), is this a problem with some sort of caching on the server side?

This would be trivial, except that given my own behavior and that of the last two submitters, I suspect we will often have situations where someone submits changes from a local collection, then immediately realizes they left something out or did it wrong, makes the changes, and resubmits, only to find that the new changes aren't reflected in the next patch submission.

For reference, the patch I tried to submit through the second (clean) browser (Firefox; the screenshots above are from Chrome) from a new IDB:
patch (8).txt

[rybesh]

OK, so when I go to review these three submitted patches:

patch 225 (rejected)
patch 226 (open)
patch 227 (open)

… they are all identical, adding a new alternative label but not changing the preferred label. They display identically in the diff view, and if I download the actual JSON patches they are identical too.

So that leads me to believe this is a client side problem—for some reason the client is incorrectly generating the patch.

Next I'll go through the exact steps you describe above and see if I can recreate this problem with my own browser.