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

Re: tarballs - why in SVN?


Am Montag 14 August 2006 16:59 schrieb Linas Žvirblis:
> Miriam Ruiz wrote:
> > What's the general idea about using a patch management system? Is it good
> > or evil?

Personally I think that patch systems don't bring any additional benefit since 
we already use svn. (And personally I'm not a very big fan of patch systems).

> I think it is good:
> * When you update the patch, you only modify a single file, and not
> bunch of files all over the place. Makes it a lot easier to see the
> changes when you are working with the source outside of SVN.

Well, we are using svn. Suboptimal for NMU's though.

> * Adding a new patch is simply a matter of throwing an appropriate file
> into an appropriate directory. More or less.

This is even easier without patch system... just modify the files and 
commit ;).

> * Things can be easily tested with only a subset of patches. This is
> especially useful in case of regressions.

However with svn you have the possibility to check out different revisions and 
see when a breakage occured.

However both patch systems or svn don't automatically make packages better. 
They are only as good as the people who work with them. Here some examples 
I've seen in the past that show that svn/patch systems can also be "misused" 
and thus won't bring much benefit:

* If I have a set of patches, I'd like to see a good description what each 
patch does, preferable in one place. This isn't guaranteed when using a patch 
system (many dpatches lack a description or have an incorrect description, 
with simple-patchsys it's a good idea to describe the patch with the name in 
the changelog, but lazyness often wins here, with svn the log entries should 
describe what has been changed but this is not always the case).

* With patch system you can do evil things like incremental patches (01_foo 
modifying file baz, 02_bar modifying baz incrementally since 01_foo was 
wrong), which are imo making things unnecessary difficult. (I've even seen a 
new upstream version being introduced as a patch once.)

* It's imho a misbelief that the use of patch systems will make upgrading to 
newer upstream versions easier. You will still have to merge the patches by 
hand, even if they apply cleanly (otherwise it's not guaranteed that they 
still do what they intend to do, or if they don't apply cleanly that bits 
from them aren't needed any longer).

* Doing svn commits that are not atomic (as in they break stuff: r1 builds 
fine, r2-r5 don't r6 builds fine again) makes tracing back individual changes 

* Committing very big hunks in one commit also makes backtracing difficult 
(the same can be said about patch-files that are very big and contain a 
number of different changes).

Anyway, I don't see a problem if we don't have a common solution wether to use 
patch systems or not. I guess it should be the choice of the persons who do 
the most packaging work for a given package.


Attachment: pgpvmxfvhJACm.pgp
Description: PGP signature

Reply to: