[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Tree conflict resolution considered harmful

From: Chris <devnullaccount_at_yahoo.se>
Date: Thu, 30 Aug 2018 06:15:58 +0000 (UTC)

Hi,

this seems to be the same issue that I reported back here (https://mail-archives.apache.org/mod_mbox/subversion-users/201804.mbox/browser). The bug fix that Stefan Sperling did for 1.10.2 has helped in some of my merges by breaking off earlier, but it is still equally slow at checking each revision. The case reported in this thread seems to have the conflict far back so I guess that's why the mitigation doesn't work here.
Unfortunately, I can't provide any more data on what happens (company-internal server), but I might be able to test run something if you have some theories on what to test.

By the way, if I remember correctly, --quiet works to stop the resolver, but --non-interactive did not.

/Chris

--------------------------------------------
On Wed, 8/29/18, Stefan Sperling <stsp_at_elego.de> wrote:

 Subject: Re: Tree conflict resolution considered harmful
 To: "Dag-Erling Smørgrav" <des_at_des.no>
 Cc: users_at_subversion.apache.org
 Date: Wednesday, August 29, 2018, 5:39 PM
 
 On Wed, Aug 29, 2018 at 12:54:43PM +0200,
 Dag-Erling Smørgrav wrote:
> I'm using
 Subversion 1.10.2 to perform a non-interactive merge with
> around 15 tree conflicts (files that exist
 on the source branch but have
> been
 deleted from the target branch).  It spends exactly two
 hours on
> each conflict before the
 connection is killed and it gives up and moves
> on to the next.  Here's an excerpt
 from ktrace showing svn's attempt to
> resolve the first conflict:
>
>  33821 svn     
 35.021454 GIO  fd 1 wrote 20 bytes

       "\rChecking r338344..."
> --
>  33821 svn   
   40.898214 GIO  fd 1 wrote 20 bytes
>        "\rChecking
 r338059..."
>  33821 svn     
 40.898328 GIO  fd 1 wrote 20 bytes

       "\rChecking r333678..."
>  33821 svn      40.898412 GIO  fd 1
 wrote 20 bytes
>       
 "\rChecking r333677..."
> --
>  33821 svn      40.900558 GIO  fd 1
 wrote 20 bytes
>       
 "\rChecking r333490..."
> --
>  33821 svn      77.091446 GIO  fd 1
 wrote 20 bytes
>       
 "\rChecking r333389..."
> --
>  33821 svn      95.000296 GIO  fd 1
 wrote 20 bytes
>       
 "\rChecking r333300..."
> --
>  33821 svn      95.001008 GIO  fd 1
 wrote 20 bytes
>       
 "\rChecking r326169..."
> --
>  33821 svn      671.067538 GIO  fd 1
 wrote 20 bytes
>       
 "\rChecking r322052..."
> --
>  33821 svn      671.337258 GIO  fd 1
 wrote 20 bytes
>       
 "\rChecking r321369..."
> --
>  33821 svn      7240.543297 GIO  fd
 2 wrote 62 bytes
>        "svn:
 warning: W210002: Network connection closed unexpectedly
>
> The third column is
 the time elapsed since the start of the process.
>
> The actual conflict
 is in r294466, which removed the file in question
> from the target branch.  The revision
 it's stuck on, r321369, only
>
 touched the svn:mergeinfo property on the current directory
 (propagated
> down from a merge higher up
 in the tree).
 
 This
 does indeed look like a problem.
 
 Unfortunately, you're not providing
 sufficient information. Ktrace tells
 us that
 time was spent somewhere. But it does not tell us what this
 time
 was actually spent on.
 
 Could you provide a shell
 script I can use to actually run the problematic
 merge myself in a working copy of the
 appropriate subdirectory of the FreeBSD
 repository? Ideally, such a script would assume
 nothing beyond a new empty
 directory and an
 'svn' binary in $PATH, and then allow me to proceed
 to the
 actual problem you were seeing.
 I'll run this script on OpenBSD, so a ksh
 compatible script would be nice.
 
 If you could take the time to
 provide that, then I will take the time to look
 into the problem. But your report so far
 doesn't provide enough material for
 me
 to chew on.
 
 Thanks!
 
 -----Inline Attachment Follows-----
 
 
Received on 2018-08-30 08:16:19 CEST

This is an archived mail posted to the Subversion Users mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.