[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Feature request (with patch)

From: Martin Burnicki
Subject: Re: Feature request (with patch)
Date: Mon, 10 Sep 2012 15:27:11 +0200
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:15.0) Gecko/20120826 Firefox/15.0 SeaMonkey/2.12

Thien-Thi Nguyen wrote:
() Martin Burnicki <address@hidden>
() Tue, 28 Aug 2012 13:17:06 +0200


    @project g:/mbglib/mbglib.pj;
    project /repository/clkdrv/linux/soft/mbgclock/mbgclock.pj;
    project g:/clkdrv/win_2k/soft/sys/mbgclock/mbgclock.pj;

    This section is located before the "desc" section of an archive file.

    The current version of GNU rcs is unable to access archive files
    which have such section since the section type is not supported, so
    Hannes Küttner, whom I've added in CC:, has made a patch which
    detects unknown sections when an archive file is read, and simply
    writes the unknown sections back when the archive file is rewritten.

    [ref mbg_rcs_unexpected_keywords.patch]

Thanks for the patch.  Although RCS hungers for users, my immediate
reaction is to regretfully point out that the comma-v syntax is frozen
(see 5.8 NEWS) in preparation for yet-unspecified "integrity features",
and thus reject it as it stands.

OK, but it's a pity since we can't use GNU RCS without this patch. It would have been easier for us if you had picked the patch up so we could have upgraded easily to newer rcs versions in the future.

That said, i now view that decision to be incomplete in that it left no
clean way to add user-defined extensions, aside from hijacking the
‘integrity’ header directly.  To DTRT, i think we need to specify that
‘integrity’ be composed of a system part and a user part, the latter
opaque (pass-through) to RCS.  These parts should be separated by some
"uncommon" byte (something in [0x01,0x1F], ideally), with the system
part coming first, and the user part (and sep byte) optional.

Some examples, w/ formfeed (ASCII 0x0C, '\f', ^L) as the sep byte:

- degenerate
   integrity @@;

- degenerate w/ user part
   integrity @^L@;

- degenerate system part, non-degenerate user part
   integrity @^L
     project g:/mbglib/mbglib.pj;
     project /repository/clkdrv/linux/soft/mbgclock/mbgclock.pj;
     project g:/clkdrv/win_2k/soft/sys/mbgclock/mbgclock.pj;

Thus, the overall change is to push user data out of the top-level, and
into the second-half of ‘integrity’, a conceptually consistent (IMHO)
location that satisfies forward compatability (at the cost of having to
manually convert some comma-v files via a sed script).

What do you think?  FYI, prior discussion on the system part is at:

    [t300 workaround]

If user data is part of ‘integrity’, t300 should pass w/o modification.

Sounds good, but doesn't help us since we actually need to share the repos with the 3rd party rcs system which adds this private header field our patch should take care of.

    Hannes Küttner has also rewritten the complete graphical user
    interface in python3, so there's also a modern front end available
    for the current version of GNU rcs.

    If someone is interested we can send them a copy of the python GUI.

Long-term RCS will provide a shared-object library (and API), but stop
short of a graphical interface.  If your code and its documentation are
available under free licenses (see <>), i
would be happy to add a link to its homepage from the RCS homepage.

Actually the python program just calls the binaries. This is very new stuff and still needs some testing, but once we have found it is stable we can make it available under a free license.


Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont

reply via email to

[Prev in Thread] Current Thread [Next in Thread]