Mail Archives: djgpp-workers/2000/12/04/20:28:57
At 09:44 PM 12/4/00 +0100, Tim Van Holder wrote:
>> ??? How does one run a local test make when the "original" dates
are
>> not available?
>Basically, you just do a checkout - the timestamps you get are the
>timestamps of the file's checkin (but they DON'T matter to cvs; you
>could run 'find -type f - exec touch {} ;' and cvs would still
function
>normally (it would simply update it's idea of each file's timestamp,
as
>kept in CVS/Entries, the next time you run 'cvs status').
>So you can just do the usual stuff (make, editing, etc). And
when >you're done, all you need to do is 'cvs diff' and voilą: instant
patch.
>djlsr203 shouldn't really come into it, unless you're trying to make
a
>2.03-to-CVS diff, in which case you're better off using 'cvs diff'
too;
>I expect there'll be a tag for the 2.03 release (checking... yup:
>v2_03)
Yes, but that assumes one *wants* to do online diffs. I just want to
be able to do local diffs to a current base. I don't really want to
re-connect to my ISP every time I need to make up a test diff. And I
don't want to checkout the files, either, because I do not feel
competent enough at cvs to be *trusted* to check anything in. I just
want to export them and generate local diffs, and then submit the patch
file to the djgpp-workers list for those who *can* be trusted to update
the real source.
And at dial-in speeds of only 50K bps or so, *nothing* I run online is
"instant". More like, go brew a pot of coffee and maybe it will be
done when you get back.
Plus, not having the original dates means that I now I have to go back
and somehow patch my changes into the "updated" sources, instead of
just being able to diff the newer sources and my tested (on a base of
v2.03) changes, in order for a locally-run make to recognize *my*
changes to the files.
<Terrific, simple instructions snipped>
>> I found out what my trouble was -- I had to add the -w option to
diff
>> to ignore CR's at EOL. The CVS files seem to all have
CRLF-delimited
>> lines, while the djlsr203.zip has (mostly) LF-delimited lines.
>> Perhaps the cvs that I'm using (the one from the cvs site for
>> windows) is producing this? Is there a way to tell cvs *not* to
use >> CRLF, but just LF? Or are the files stored in the repository
that
>> way?
>No, it's an unfortunate artifact of the NT port of CVS (and I believe
>the DJGPP package I made also had this problem): it writes text files
>as text (surprise, surprise), resulting in CRLF on DOS. 'cvs diff'
>doesn't care though - when diffing a text file, it will read data in
>text mode as well.
Well, it does explain the problem. I had already started to make up a
shell script to copy the date of the files from the v2.03 source zip to
the ones with the same name in the exported copy in my sandbox area,
but I realized after starting on it that it was a mistake, since any of
those files may be the ones that were updated since v2.03. With the -w
option, at least a local diff sees only the files that have really
changed since v2.03. I could (and maybe I will) update the script to
run dtou on all of the files in the sandbox, which would at least
eliminate the use of -w.
>Hope this helps,
Tim, it certainly does. Thank you for those excellent instructions,
they're *almost* enough to make me feel I understand what I'm
doing. The failure is mine, not yours.
<*Sigh*> And I thought I was almost done with these changes...
---------------------------------------------------------
Peter J. Farley III (pjfarley AT dorsai DOT org OR
pjfarley AT banet DOT net)
- Raw text -