in order to determine if the directory of a file being added is already under version control, the code looks for a ".svn" in the directory alongside the file. SVN 1.7 no longer uses ".svn" in every directory, so the check fails, the code starts crawling upwards, and ends up trying to do things it shouldn't.
What's the right way of doing this check? We could run "svn info"...
that might be it. I am attempting to repro/fix it right now (suppose I might as well assign this to me, then :).
Yep, AFAIK "svn info" is the right way to go, and is backwards-compatible. Too bad it's a lot slower. However, I think it can get you to the root in one shot instead of needing to recurse.
Created attachment 142586 [details] Patch
Comment on attachment 142586 [details] Patch Could we check for .svn and if it doesn't exist, run svn info? That would optimize for the common case until svn 1.7 usage grows.
(In reply to comment #5) > (From update of attachment 142586 [details]) > Could we check for .svn and if it doesn't exist, run svn info? That would optimize for the common case until svn 1.7 usage grows. Sure. It's not clear to me that 'svn info' is particularly slow - I think it's a local call - but that's a simple enough optimization.
Created attachment 142595 [details] add shortcut for svn < 1.7
Committed r117526: <http://trac.webkit.org/changeset/117526>
Re-opened since this is blocked by 86801
Rolled out ( :( ) because the Chromium Mac 10.5 layout test canaries broke. Here's a sample output: http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Mac10.5/builds/11591/steps/webkit_lint/logs/stdio
Comment on attachment 142595 [details] add shortcut for svn < 1.7 Clearing ? from the patch that had been landed and rolled out.
Created attachment 142794 [details] Patch
Comment on attachment 142794 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=142794&action=review OK. > Tools/ChangeLog:9 > + Re-land the change in r117526 with a fix to maintain > + compatibility with SVN 1.4.4 (yay Leopard!); turns out Really? Leopard be damned, no?
(In reply to comment #13) > > Tools/ChangeLog:9 > > + Re-land the change in r117526 with a fix to maintain > > + compatibility with SVN 1.4.4 (yay Leopard!); turns out > > Really? Leopard be damned, no? Chromium still has Leopard bots.
Committed r117643: <http://trac.webkit.org/changeset/117643>
(In reply to comment #14) > (In reply to comment #13) > > > Tools/ChangeLog:9 > > > + Re-land the change in r117526 with a fix to maintain > > > + compatibility with SVN 1.4.4 (yay Leopard!); turns out > > > > Really? Leopard be damned, no? > > Chromium still has Leopard bots. Current plan is to kill them sometime this Summer.