In other news, I looked into the cause --- tried to make datasource_get_next_token() do one more loop in the place where currently it does 'return if at_start_of_suffix()' --- but that didn't fix the truncation...
In the meantime, I tweaked a test to make it XFail (r1068798). From a quick glance it seemed to be relevant, but in second thought I'll admit I didn't study the test thoroughly before making the patch. Daniel Daniel Shahaf wrote on Wed, Feb 09, 2011 at 11:10:43 +0200: > Johan Corveleyn wrote on Wed, Feb 09, 2011 at 08:42:20 +0100: > > On Wed, Feb 9, 2011 at 4:54 AM, Daniel Shahaf <d...@daniel.shahaf.name> > > wrote: > > > The experimental code is svn_diff_diff4_2(); AFAIK svn_diff_diff4() is > > > as stable as ever. > > > > I have no objections on adding such an annotation. However, from where > > I'm sitting, svn_diff_diff4() is/was just as under-exercised as > > svn_diff_diff4_2(), i.e. no known callers, only one unit test (thanks > > for adding that test, BTW). Of course it's into the codebase a lot > > longer than *_2, but it has never had any core code calling it, and no > > unit tests (so could have been broken by any number of commits after > > its inception till now). > > > > In other words, svn_diff_diff4() is as experimental as svn_diff_diff4_2(). > Agreed. > > IMO the issue boils down to representation: if we haven't tested a given > piece of code (it has no callers and a surfacial unit test), then we > shouldn't represent to API consumers otherwise. > > Daniel > (and yes, I respect all the work you've been doing; my opinion of > diff4_2() is orthogonal to that) > > > -- > > Johan
