[cvsnt] Re: Keyword 'properties' breaks rcs keyword ordering

[Matt Schuckmann]

The viewcvs python code actually has several methods for parsing the RCS
file.

The most often used is to ask cvs to produce an rlog but it does look like
there was an atempt to parse the RCS file directly, it looked like it was
experimental and not well used.

I'm not sure what cvsgraph does, it sounds like it parces the rcs file
directly.

Matt S.

"Tony Hoyle" <tmh at nodomain.org> wrote in message
news:d8ks00$fm5$1 at paris.nodomain.org...
> David Somers wrote:
>
> > Which tends to indicate it is that pesky 'properties' in the wrong place
> > error. Looks like somebody will have to hack the viewcvs code (don't
look at
> > me, I don't really do python)... either that, or write some code to
repair
> > the affected files in a cvsnt repository.
> >
> My understanding is that viewcvs calls 'cvs rcsfile' directly.  I'm not
> sure why it would ever parse the RCS files like that.
>
> I can move properties (plus stop it appearing when it shouldn't) but it
> won't really help parse something like:
>
> properties
> notes:@fixes a bug in the widgets code; to test.@
> to_test:@development team 3@;
>
> ..there's multiple @ strings plus a semicolon in there.. it doesn't
> conform to the 'newphrase' syntax at all.
>
> CVSNT is also rather more liberal with what it puts in the files these
> days than RCS ever was (relaxed tag naming rules etc.)
>
> Tony