How to make 'svn log' ignore property changes?

Question

'svn log' shows the log of: 1) files whose content has changed AND/OR 2) files whose properties have changed.

Is there a way to show only files for which case 1 applies?

Answer 1

Doesn't look like it: SVN log in the Subversion manual

Maybe filter them out afterwards, e.g. using a shell script? What system are you on?

Answer 2

Hi zr,

Even before searching around I came to the conclusion that this would require taking each revision in the log and passing it to `svn diff' to see just what changed in the revision. Then I found this other SO question: how to detect modified properties using SVN log

There is svnlog.py which includes filtering options for authors and paths, but does not include any diff integration. You might be able to hack that.

I hope this helps.

Thank you,
Zachary

Answer 3

'svn log' shows the log of: 1) files whose content has changed AND/OR 2) files whose properties have changed.

As you know, svn log's output is organized by commit log messages rather than by files. If you use verbose mode (-v), it will show the files ("paths") associated with each log entry. However some of those paths can be outside of the requested target (default: your current directory). Do you want the results to include those external paths as well? I guess taking your question at face value, you're just asking for filtering, so yes you would want those external paths if they represent a content change to files.

Here is a solution. It may be slow, but I tested it and it works in cygwin on Windows 7. (Don't forget, make sure your scripts have unixy line endings! with dos2unix if necessary)

It's really only one long line, except for an external sed script (which a real hacker could put on the command line but life is short to be messing with command-line escaping):

#!/bin/sh

# These paths are set up for cygwin
SED=/bin/sed
SORT=/bin/sort
UNIQ=/bin/uniq
XARGS=/bin/xargs
GREP=/bin/grep
SVN=svn

# Add desired log options here
LOG_OPTIONS=-v
SINCE_REV=10800
SED_SCRIPT=/cygdrive/c/temp/get-paths.sed
# Make sure you edit the sed script referenced above
# to set the base URL of your repository.

# 1) generate the list of log messages, including affected paths (svn log -v)
# 2) process out those paths (sed)
# 3) eliminate duplicates (sort | uniq)
# 4) get the change history for each of those paths (svn diff)
# 5) filter out the ones that involve only property changes (sed)
# 6) eliminate duplicates again (sort | uniq)
$SVN log $LOG_OPTIONS | $SED -n -f $SED_SCRIPT | $SORT | $UNIQ \
 | $XARGS -n20 -I PATHS $SVN diff -r $SINCE_REV --summarize PATHS 2> error.log \
 | $SED -n 's/^[^ ].... *//p' | $SORT | $UNIQ

Here is the external sed script. Be sure to change the svn repository base URL to the right base URL for your repository. I.e. the beginning part of the svn URL that is not output by svn log -v.

 # sed script to output all lines between
 # /^Changed paths:$/ and /^$/, exclusive.
 # Also removes first 5 columns (status) and replaces them with svn repository base url.

 /^Changed paths:$/,/^$/ {
   /^Changed paths:$/b
   /^$/b
   s|^.....|https://svn.myrepo.org/prefix|
   s/ (from .*)$//
   p
}

The script will output some error messages to error.log, primarily "path not found", which I believe is for files that used to be present in the repository but have been moved (renamed) or deleted.

Does this meet your requirements?

Credit to Michael Augustin at this page for ideas about grepping the output of svn diff to remove property-only changes.

P.S. This other page seems to ask the same question, but there's no full answer there.

P.P.S. Edited the above bash script to add an extra | sort | uniq to the end of the pipeline, since I have seen duplicates come out there. Though I don't see why they would occur.