cvsexportcommit and cvsimport's -k option

cvsexportcommit and cvsimport's -k option

[Lars Alexander Noschinski]

[Please CC me on answers, as I'm not subscribed.]

Hello!

Im using git to work on an CVS project. My work cycle is as follows:

... hack away ...
git-cvsimport -i -r cvs -kk -v -d $CVSREPOS
git-rebase cvs # If work was done in the meantime
git-cvsexportcommit -cpuv $COMMIT # Multiple times, if necessary
git-cvsimport -i -r cvs -kk -v -d $CVSREPOS
git-checkout -b master cvs # commit ids are changed now

As far as I am concerned, this works fairly well, accept when for any reason
a line containing a CVS keyword (e.g. $Id$) is contained in a changeset. In
this case, patching fails, and I have to investigate why it fails, eventually
just copying the old version from the cvs repository.

Would it be possible to add an option to cvsexportcommit to detect such
killed keywords and ignore them?


Greetings,
   Lars.

Re: cvsexportcommit and cvsimport's -k option

[Robin Rosenberg]

tisdagen den 29 april 2008 19.15.06 skrev Lars Alexander Noschinski:
> ... hack away ...
> git-cvsimport -i -r cvs -kk -v -d $CVSREPOS
> git-rebase cvs # If work was done in the meantime
> git-cvsexportcommit -cpuv $COMMIT # Multiple times, if necessary
> git-cvsimport -i -r cvs -kk -v -d $CVSREPOS
> git-checkout -b master cvs # commit ids are changed now
>
> As far as I am concerned, this works fairly well, accept when for any
> reason a line containing a CVS keyword (e.g. $Id$) is contained in a
> changeset. In this case, patching fails, and I have to investigate why it
> fails, eventually just copying the old version from the cvs repository.
>
> Would it be possible to add an option to cvsexportcommit to detect such
> killed keywords and ignore them?

Sure, but is it necessary? If the cvs checkout you use is checked out
with -kk there shouldn't be any problem.

If not, start by thinking up a few testcase for inclusion in 
t9200-git-cvsexportcommit.sh that illustrates the problem. Then we can
change cvsexporcommit until the new tests pass.

-- robin

Re: cvsexportcommit and cvsimport's -k option

[Lars Noschinski]

Hello!

* Robin Rosenberg <robin.rosenberg.lists@dewire.com> [08-04-29 22:56]:
>tisdagen den 29 april 2008 19.15.06 skrev Lars Alexander Noschinski:
>> ... hack away ...
>> git-cvsimport -i -r cvs -kk -v -d $CVSREPOS
>> git-rebase cvs # If work was done in the meantime
>> git-cvsexportcommit -cpuv $COMMIT # Multiple times, if necessary
>> git-cvsimport -i -r cvs -kk -v -d $CVSREPOS
>> git-checkout -b master cvs # commit ids are changed now
>>
>> As far as I am concerned, this works fairly well, accept when for any
>> reason a line containing a CVS keyword (e.g. $Id$) is contained in a
>> changeset. In this case, patching fails, and I have to investigate why it
>> fails, eventually just copying the old version from the cvs repository.
>>
>> Would it be possible to add an option to cvsexportcommit to detect such
>> killed keywords and ignore them?
>
>Sure, but is it necessary? If the cvs checkout you use is checked out
>with -kk there shouldn't be any problem.

Errm, now as you mention it ... Even it is clearly mentioned in
git-cvsimport(1), it did not realize, that -k just invokes cvs with -kk.
I think this is a perfectly fine solution.

>If not, start by thinking up a few testcase for inclusion in
>t9200-git-cvsexportcommit.sh that illustrates the problem. Then we can
>change cvsexporcommit until the new tests pass.

As cvs can do this for us, I think it is not necessary to introduce
another source of potential errors.


Thanks for your quick answer and sorry for me being dumb,
     Lars