When using a remote SVN repository, we are including too many commits in our history. This results in us computing the wrong identifiers for branch commits.
<rdar://problem/71953465>
Not sure there is much we could do in testing to prevent this, other than having a set of tests that compare the results on an actual checkout with those of an actual remote, which is something we have traditionally tried to avoid.
Created attachment 415352 [details] Patch
From a Slack conversation explaining the motivation: The request being modified here was reverse engineered from inspecting packets sent by the svn binary. I had stripped a few pieces out that I thought we didn't need so we had less data being transferred, but it appears I stripped out a piece we did need (that is, the "<S:path></S:path>" bit)
Created attachment 415355 [details] Patch for landing
Committed r270412: <https://trac.webkit.org/changeset/270412> All reviewed patches have been landed. Closing bug and clearing flags on attachment 415355 [details].