On Monday 31 January 2011 08:01 PM, C. Michael Pilato wrote:
On 01/31/2011 04:48 AM, Arwin Arni wrote:
Hi All,
This patch revs subversion/libsvn_subr/config.c:svn_config_read to
svn_config_read2 to accept a *boolean* called *section_names_case_sensitive*.
When TRUE, the section names are populated case sensitively in the config.
This was done mainly for the case insensitivity issue in authz file parsing.
(Issue #3781<http://subversion.tigris.org/issues/show_bug.cgi?id=3781> ->
http://subversion.tigris.org/issues/show_bug.cgi?id=3781 )
Here is a relevant thread: http://svn.haxx.se/dev/archive-2011-01/0441.shtml
Currently, this patch will only treat section names case sensitively.
Should I add parameters for options / values / options& values to decide
the case or shall I use a common parameter for all(sections, options& values) ?
We generally try to avoid premature over-engineering, and at the moment, we
have no known need for the additional changes you speak of. Introducing
codepaths that we don't exercise can lead to surprise failures later when we
do.
But I am excercising it!
In libsvn_repos/authz.c. I'm using it to make the config population
case-sensitive. (Issue #3781)
All other callers pass FALSE to that parameter (unless they suddenly
require case-sensitivity).
I hope I've made things a little clearer. :)
That said, I suppose if you're able to give sufficient unit test
coverage, that's enough "exercise". I could go either way on this one.
(Sorry... I don't suppose that was very helpful, was it?)
A colleague of mine has already started working on a test case for this.
Looking forward to hear more about this..
Thanks and regards,
Arwin Arni
