RE: svn commit: r1607306 - /subversion/trunk/subversion/svnadmin/svnadmin.c

[Bert Huijben <be...@qqmail.nl>]

> -----Original Message-----
> From: stefan2@apache.org [mailto:stefan2@apache.org]
> Sent: woensdag 2 juli 2014 12:06
> To: commits@subversion.apache.org
> Subject: svn commit: r1607306 -
> /subversion/trunk/subversion/svnadmin/svnadmin.c
> 
> Author: stefan2
> Date: Wed Jul  2 10:05:46 2014
> New Revision: 1607306
> 
> URL: http://svn.apache.org/r1607306
> Log:
> Eliminate FSFS format 7 CPU overhead from svnadmin.
> 
> This introduces a simply threshold value for the cache sizes above
> which we will enable block-read.
> 
> While svnadmin is the ideal application for "block-read" (read all
> repo data somewhat linearly and use virtually all of it), the very
> small default cache sizes are simply not enough to benefit from it.
> As a result, hit rates drop even further.
> 
> For larger caches, the feature equalizes most of the structural
> disadvantage that packed f7 repos have over f6 (f6 is in revision
> order and svnadmin is the only tool to read it in that ordering).
> For smaller caches, only meta data and certain temporary objects
> get cache while we heavily rely on the OS disk cache to provide
> us with data.  This is equally true for all FSFS formats.

As you probably found out by now (on all buildbots):
 [[[
W: Unexpected error while running 'svnadmin verify'.
W: EXPECTED STDOUT (regexp):
W: | .*Verifying.*metadata.*
W: | .*Verified revision 0.
W: | .*Verified revision 1.
W: | .*Error verifying revision 2.
W: | .*Error verifying revision 3.
W: | .*
W: | .*Summary.*
W: | .*r2: E200014:.*
W: | .*meta data at offset.*
W: | .*expected.*
W: | .*actual.*
W: | .*
W: | .*r3: E200014:.*
W: | .*meta data at offset.*
W: | .*expected.*
W: | .*actual.*
W: | .*
W: ACTUAL STDOUT:
W: | * Verifying metadata at revision 0 ...
W: | * Verified revision 0.
W: | * Verified revision 1.
W: | * Error verifying revision 2.
W: | * Error verifying revision 3.
W: | 
W: | -----Summary of corrupt revisions-----
W: |     r2: E160004: Corrupt node-revision '0-2.0.r2/3'
W: |     r2: E160004: Found malformed header 'inserting junk to corrupt the rev' in revision file
W: |     r3: E160004: Corrupt node-revision '0-2.0.r2/3'
W: |     r3: E160004: Found malformed header 'inserting junk to corrupt the rev' in revision file
W: CWD: E:\svn-ra\tests\subversion\tests\cmdline
W: EXCEPTION: SVNLineUnequal
Traceback (most recent call last):
  File "D:\ra\svn-ra\build\subversion\tests\cmdline\svntest\main.py", line 1634, in run
    rc = self.pred.run(sandbox)
  File "D:\ra\svn-ra\build\subversion\tests\cmdline\svntest\testcase.py", line 254, in run
    return self._delegate.run(sandbox)
  File "D:\ra\svn-ra\build\subversion\tests\cmdline\svntest\testcase.py", line 176, in run
    return self.func(sandbox)
  File "D:\ra\svn-ra\build\subversion\tests\cmdline\svnadmin_tests.py", line 2116, in verify_keep_going
    output, errput, exp_out, exp_err):
  File "D:\ra\svn-ra\build\subversion\tests\cmdline\svntest\verify.py", line 456, in verify_outputs
    compare_and_display_lines(message, label, expected, actual, raisable)
  File "D:\ra\svn-ra\build\subversion\tests\cmdline\svntest\verify.py", line 429, in compare_and_display_lines
    raise raisable
SVNLineUnequal
FAIL:  svnadmin_tests.py 32: svnadmin verify --keep-going test
]]]

	Bert

Re: svn commit: r1607306 - /subversion/trunk/subversion/svnadmin/svnadmin.c

[Stefan Fuhrmann <st...@wandisco.com>]

[ maybe this gets through the spam filter ?! ]

Sorry! Fixed in r1607377.

-- Stefan^2.