Stefan Küng wrote:
>On 10/19/05, Peter McNab <mcnab_p@melbpc.org.au> wrote:
>
>
>This might be strangely related to to an incomplete commit of a tree
>full of changes very recently.
>One file deep in the WC folder tree failed to commit and remained
>showing as changed even after all the other files committed
>successfully. It was a binary file (dll) that replaced an earlier
>version. To be sure it wasn't a TSVNCache issue I killed the cache.
>After the new cache instance had completed it trolling the file
>continued to show as changed. The file could not be committed either
>until I did an update and then it committed OK. The particular folder is
>one that is only ever changed when a new version of the vendor code is
>release and never produced by local compilation. I have no reason to
>consider that that folder alone was not fully up to date unless
>something similar to what is described above has occurred.
>
>
>
Unsure if the path length comment was also directed at my comment about
the failed complete commit so providing it anyway.
>How long are the paths? If you have paths longer than 255 chars, then
>it won't work.
>
>
>
OK don't think it's an overlong path issue for my failed full commit..
The path was
http://192.168.1.20/svn/FastMM/Trunk/Replacement BorlndMM
DLL/Precompiled/for Applications/Performance
>Please try the Subversion CL client. TSVN uses the Subversion API for
>all these features, so if TSVN fails here, it has to be Subversions
>fault.
>
>
The situation I experienced is not repeatable because of the nature of
the inflowing info.
However there have been a couple of times over the past few weeks where
I have had to go back and grab a file I thought I must have missed in a
merge or branch/tag creation and had put it down to my age related
forgetfulness. However I'm getting the worrying feeling that Subversion
on Windows (well possibly the libraries incorporated within TSVN) have
not crawled (recursed) the folder tree completely, so missing the odd
file, more at random than repeatably. Sometimes these events don't
reveal themselves immediately, but sometime later when a compile fails,
so the trail of action has already blurred.
The above is only a gut feeling, not a certainty and not something for
devs to go chasing fairies just yet.
When I get back into coding rather than cataloging I will keep a more
alert eye out for anomalies.
Peter
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tortoisesvn.tigris.org
For additional commands, e-mail: users-help@tortoisesvn.tigris.org
--
Received on Wed Oct 19 13:33:23 2005