Package: darcs
Version: 2.0.0-5
Severity: minor

Someone recorded a monolithic patch, and I wanted to back out a couple
of hunks.  So I did "darcs rollback", removed the first bad hunk.  Now
"darcs rollback" won't allow me to re-rollback the same patch (but a
different hunk within it).

-- System Information:
Debian Release: lenny/sid
  APT prefers testing
  APT policy: (990, 'testing')
Architecture: i386 (i686)

Kernel: Linux 2.6.25-2-686 (SMP w/1 CPU core)
Locale: LANG=en_AU.utf8, LC_CTYPE=en_AU.utf8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash

Versions of packages darcs depends on:
ii  libc6                  2.7-10            GNU C Library: Shared libraries
ii  libcurl3               7.18.2-5          Multi-protocol file transfer libra
ii  libgmp3c2              2:4.2.2+dfsg-3    Multiprecision arithmetic library
ii  libncurses5            5.6+20080308-1    Shared libraries for terminal hand
ii  zlib1g                 1:1.2.3.3.dfsg-12 compression library - runtime

Versions of packages darcs recommends:
ii  msmtp-mta [mail-transport-age 1.4.13-1   light SMTP client with support for

-- no debconf information

Hmm, looks like you would have to unpull your rollback patch and do it again
(rolling back both hunks in a single patch)

On Thu, Jul 24, 2008 at 10:01:35AM -0000, Eric Kow wrote:
> Eric Kow <eric.kow@gmail.com> added the comment:
> 
> Hmm, looks like you would have to unpull your rollback patch and do it again
> (rolling back both hunks in a single patch)

Yup.  I find that counter-intuitive, hence the bug report.  OTOH it
seems reasonable to me that for the darcs team to just say "you
should've sent the monolithic patch back and asked the guy to record
discrete patches".

This has been resolved since darcs 2.

Just commenting on this: the darcs2 rollback is *really* nice once you get the
hang of it.

The trick is to understand that this is not by any means the old rollback.  No
invert patch bits are being flipped or anything silly like that.  Rollback is
merely a command for automatically generating new patches...

More to the point: if you have
patch A then
patch B

You may notice that rollback sometimes refuses to rollback B if you do not
select A.  The trick is simply to accept the rolling back of A in order to get
at B... then when darcs ask if you want to rollback individual changes in A, you
say /no/ and just select the changes in B that you want to rollback.  The reason
why I'm writing this is that some users may instinctively balk at the idea of
selecting A as one of the patches to rollback.

The reason why this works is that not all the changes in A depend on B, so you
can easily rollback just the changes in B you want.

I'm not sure how to overcome that initial resistance.  I wonder if something
somewhere needs to be explained better, or if we need to give more examples.  It
seems like part of learning to use rollback is learning to trust it  to do the
right thing :-)

As usual, you can always check to make sure that your rollback does the right
thing by inspecting its contents (darcs changes --last=1) and unpull it.