Re: svn_client_diff4 and valid relative path throws error even though it shouldn't

On Nov 23, 2007 3:14 PM, Stefan Küng <tortoisesvn@gmail.com> wrote:
>
> Mark Phippard wrote:
> > On Nov 23, 2007 2:54 PM, Stefan Küng <tortoisesvn@gmail.com> wrote:
> >> Just updated to r27993 of the svn trunk. The new param relative_to_dir
> >> in svn_client_diff4() is nice, but it doesn't work in some situations:
> >>
> >> a diff between a working copy file and it's base throws an error like this:
> >> Path 'C:\Users\kueng\AppData\Local\Temp/svndiff.tmp' must be an
> >> immediate child of the directory
> >> 'D:/Development/SVN/TortoiseSVN/ext/Subversion'
> >>
> >> The problem: I didn't specify any temp file path. That path comes from
> >> the fact that the diff first 'translates' the base file (line endings,
> >> maybe keywords, ...), and by doing so creates that temporary file.
> >>
> >> I would expect that since it was the svn function itself that created
> >> that temp file, it shouldn't error out but check the original path instead.
> >>
> >> To reproduce:
> >> * modify the file subversion\libsvn_subr\config_win.c
> >> * call svn_client_diff4(path, REVWC, path, REVBASE,
> >> 'subversion\libsvn_subr', ...)
> >
> > Stefan,
> >
> > Take a look at the latest patch in Issue#: 2723
> >
> > That seems to fix it for me.
>
> Bingo! That fixes the problem.
>
> I wonder though: why do full committers attach a patch to an issue
> instead of just committing it? When I discovered the problem, the first
> thing I did was updating to the latest HEAD revision to see if it might
> have been fixed already. Even though I knew about the issue with the
> 'absolute paths in diff' because I saw the mails about this, it never
> occurred to me to check the issue for a patch - I assumed the issue was
> fixed when the new API was introduced.
>
> At least now I know that I have to check the issues for patches first ;)

Jeremy (the one who worked on this issue) is not a committer. I could
have committed the patch but I wanted someone else to look at it based
on the comments I added to the issue. I am also not a C programmer so
I would only be committing based on the fact that it worked for my
tests as opposed to things like coding standards and proper API usage
etc.

In other words, if this patch had been made by a full committer it
likely would have just been committed.

-- 
Thanks
Mark Phippard
http://markphip.blogspot.com/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org