[PATCH 0 of 9] convert/gnuarch: fixes and improvements

Edouard Gomez ed.gomez at free.fr
Sat Jan 3 19:37:09 CST 2009

convert/gnuarch: fixes and improvements

I wrote a python script in 2005[1] that could successfully convert my xvidcore
gnu arch archives to mercurial. As a matter of fact, xmas hollidays is 
always a good time for digging into his old stuff... i wanted to check how
the current convert gnuarch support was doing against my old script. Ouch...

The current gnuarch backend has some shortcomings that were preventing me
from converting my gnu arch archives. So after a bit of work, here is
a patch series that address most of the errors and shortcomings.

Patches in the 'fix' category:
 - convert/gnuarch: remove unused code
 - convert/gnuarch: fix cat-log parsing
 - convert/gnuarch: set prefered locale for str conversions

Patches in the improvement category:
 - convert/gnuarch: robustify cat-log retrieval
 - convert/gnuarch: keep track of original revision in extra headers

Patches in the "improve history scanning thus allow more complete conversions":
 - convert/gnuarch: use fully qualified revisions
 - convert/gnuarch: parse continuation-of revisions in gnuarch source
 - convert/gnuarch: retrieve known archive names list
 - convert/gnuarch: follow continuation-of revisions

All patches have a longer explanation of what they do/fix.

I could test the changes with latest GNU Arch tla client. I could not
find any bazaar 1.x tarball or debian package. So result tests with baz
or an url to baz sources would be welcome.

Please apply to crew.

PS: I have some other ideas to improve the backend, but i lack time to explore
them. The one that would bring the more improvement is:
 - parse merge information to discover more tree revisions and have more
   accurate parent information

[1] http://ed.gomez.free.fr/vrac/arch2hg.py

Edouard Gomez

More information about the Mercurial-devel mailing list