Hi Patrick,
I'm not sure who the notepad users are. Certainly, not me. I'm not
sure where notepad is on my Windows machine.
We do have a choice. We can leave things as they are and not get too
upset when developers inadvertently mess things up by accidentally
changing the line ending style on their client copy when checking in a
changed file.
Or, we can regiment the files to native, making it much more unlikely
that developers will inadvertently change the line endings.
We do know that all as-is files with UNIX (or was it DOS?) line endings
can have their eol-style attribute changed without recording a change in
every line. One of the files in my experiment yesterday did just that.
So that change alone applied across the code base might reduce the
problem children by half, from 40% of the files to 20%.
It is also possible that we can hack the other 20% by some tricky moves
that we can discover through some more experiments. Finally, maybe
there are some scripts that can be run on the repository that would
effect the change without messing with the history.
David
Patrick Linskey wrote:
On the downside, the file that was in DOS line endings was converted on
the server and recorded in the change notice as the entire file changing.
I really don't like the idea of losing all this easy history. Is this
line-ending stuff really a problem? Don't most modern text editors
support multiple line endings?
It seems a waste to lose all that data just to make things easier for
Notepad users....
-Patrick
On Feb 7, 2008 6:09 AM, David Ezzio <[EMAIL PROTECTED]> wrote:
Hi,
Based on an experiment, it looks like the regimentation of existing
files to native will be easy to perform.
In r619411, I changed two files from as-is to native. One was present
on my client with Unix line endings and one was present with DOS line
endings. I changed the eol property for both files to native and
committed. When updated from the server, they both came down with DOS
line endings, the appropriate native translation for my Windows machine.
On the downside, the file that was in DOS line endings was converted on
the server and recorded in the change notice as the entire file changing.
I propose that I find an hour or two next week and regiment all files
with a non-native setting to the native setting (about 60% of the code
base).
David
David Jencks wrote:
I think infra strongly suggest everyone uses these svn client settings
to avoid most of this kind of problem:
http://www.apache.org/dev/svn-eol-style.txt
I think there are scripts to help normalize stuff that has strayed from
these recommendations but I'm not sure where they are.
thanks
david jencks
On Feb 6, 2008, at 9:09 AM, Craig L Russell wrote:
Hi David,
Thanks for volunteering to do this.
I recall we discussed whether to use LF or native. But I don't recall
the outcome.
The reason to use LF is for Windows users who use unix tools. The
reason to use native is for Windows users who use native tools
(Notepad). I personally don't care, except that my svn preferences
file is set up to use LF for new files (for another project).
Craig
On Feb 6, 2008, at 8:41 AM, David Ezzio wrote:
Hi,
As I understand it, files within the SVN repository are either text
or binary, and if text, they have as a property one of the five SVN
settings: CR (Mac), LF (Unix), CRLF (DOS), native (convert to/from
client platform) or as-is (no conversion, no regimentation).
Currently, the OpenJPA repository has text files for three of these
settings. Most are as-is, a few are LF, and the remainder are native.
As I understand it, native is the preferred attribute for text files
as SVN will take care to convert to and from the client's platform
preference upon update and commit.
I believe it would be a relatively simple matter for me to convert
all of the files to one agreed upon format, anytime that we'd care to
do so.
Thoughts?
David
Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:[EMAIL PROTECTED]
P.S. A good JDO? O, Gasp!
