On Wed, 21 Nov 2007, Stephen Butler wrote:
> Hi folks,
> In svn_wc_conflict_description_t there is a member named 'access', a
> pointer to svn_wc_adm_access_t. It seems out of place in a struct of
> pointers to const. (subversion/libsvn_wc/svn_wc.h, line 1094)
> A full text search of the Subversion source shows that this field is
> "write-only": found only on the left side of assignments (e.g., in the
> interactive conflict resolution callbacks) and never passed as an
> Must it remain for binary compatibility, or should I submit a patch to
> get rid of the dead code?
Steve, because svn_wc_conflict_description_t is new for 1.5, we're
free to change it up until the final release without concerning
ourselves with compat (though, we can impact third-party early
adopters like TortoiseSVN and Subclipse).
Before talking about removing it, we need to understand why Ben added
it in the first place. My guess is that it's so that external library
consumers can acquire the write-lock for the conflicted area of the WC
-- which has already been write-locked by a merge/update/switch
process -- in their conflict resolution callbacks. If I'm correct, we
probably shouldn't remove it (but might want to better-document it).
Received on Wed Nov 21 18:25:53 2007
- application/pgp-signature attachment: stored