On Fri, Jul 26, 2024 at 7:08 PM Daniel Sahlberg <daniel.l.sahlb...@gmail.com> wrote: > Den tors 25 juli 2024 kl 06:44 skrev zongganli(李宗淦) <zongga...@tencent.com>: >> >> Hello, >> >> I just found that in revision 1303856 >> https://svn.apache.org/viewvc?view=revision&revision=1303856, it sets the >> send_deltas in svnlook to TRUE. >> Actually, on our svn server, we set send_deltas=FALSE, as our repos are so >> large, which takes a long time to run svnlook changed. For example, in a >> 30GB+ repo, it takes more than 20 minutes, but without send_deltas, it is >> less than 1s. >> So I wonder if there is any risk to run svnlook changed with >> send_deltas=FALSE. > > > Hi, > > Moving the discussion to dev@ since I think it better fits that audience. > > This is a very interesting case. As far as I understand, and I admit I havn't > researched exhaustively, r1293375 modified how changed properties were > reported. This caused issues for svnlook (and the ruby test suite) requiring > the changes in r1303856. Then further issues were found and r1293375 was > reverted completely in r1482436. The reverted code seems to still be around > today. I can't see if there was any other changes later on that were supposed > to accomplish the same thing in a different way. > > It seems like reverting r1303856 may be a good idea - maybe excluding a few > tests.
Interesting indeed. The only risk I can see is if there was a released server version in between r1293375 and r1482436, which is being talked to by a new svnlook with reverted r1303856. But then ... svnlook is a server-side tool, so always matches the build of the server, right? So, indeed it seems safe to me to revert r1303856 (since the reason for its introduction is no longer there). -- Johan
