Re: Strange diff behavior on branch

[Paul Sander]

>--- Forwarded mail from address@hidden

>[ On Thursday, July 31, 2003 at 07:47:09 (-0700), Mark D. Baushke wrote: ]
>> Subject: Re: Strange diff behavior on branch 
>>
>> I think that the -D "timestamp" is documented as doing timestamps on the
>> trunk rather than a branch at present. The "right thing" to do is probably
>> to either extend the -r syntax to take -r branch:timestamp or add a new
>> flag like -j branch:timestamp as the FreeBSD folks did.

>I really don't like the idea of using "cvs [r]diff -j branch:date".
>There's way too much room for confusion.

>I don't really like the idea of "cvs [r]diff -r branch:date" either,
>though if it were the only alternative it would be the better choice.

>I think extending the '-D' option to allow branch would be OK.

>But I don't think any of the above are necessary.....

Perhaps not, but storing module/branch/timestamp triples can save a lot
of time tagging and is equally correct.

>It seems to me that it would be most logical and most elegant to simply
>use the combination of '-r' and '-D' when a date on a particular branch
>is desired:

>       cvs [r]diff  -r branch-1 -D date-1  -r branch-2 -D date-2

>A usage syntax check should exit with an error if the operand(s) to '-r'
>is(are) not a branch specifier whenever '-D' is also used.


>BTW, I never really liked the idea of using the order of two otherwise
>identical command-line flags (with hopefully different operands) to
>specify the order of the diff parameters, but I guess it's way too late
>to fix that bug now.  :-)  It should have been done with a single flag
>and a two-value operand:  -r rev1,rev2

Actually, it should have been done with separate ancestor and contributor
arguments:  -a rev1 -r rev2 *or* -a rev1 -j rev2

>--- End of forwarded message from address@hidden