It looks like there is some problems with darcs replace when
token-chars contains odd characters. darcs replace exits with 0 but
nothing is shown with whatis. The previous replace changes are lost,
too. The darcs versions tested are home-brewn darcs-1.0.7 and
statically linked 1.0.8 downloaded from the net.

A shell script demonstrating problem is attached.   Here is an example run:

$ ./darcs-bug-replace
+ darcs add foo.c foo.xml
+ darcs record --all -mAdded foo.c foo.xml
Recording changes in "foo.c" "foo.xml":

Finished recording patch 'Added'
+ darcs replace foo fuu foo.c
+ darcs w
{
replace ./foo.c [A-Za-z_0-9] foo fuu
}
+ darcs replace X XX foo.xml --token-chars '[a-zA-Z_0-9$]'
+ darcs w
{
replace ./foo.c [A-Za-z_0-9] foo fuu
replace ./foo.xml [a-zA-Z_0-9$] X XX
}
+ darcs replace ' X ' ' XX ' --token-chars '[ X]' foo.xml
+ darcs w
No changes!
Exit 1

The expected results would be that the third darcs either adds the
replace change or complains about weird token-chars then exits with
non-zero value. Previous replaces should not be lost.

Attachments

• darcs-bug-replace

On Sat, Aug 05, 2006 at 05:23:52PM +0000, Pekka Pessi wrote:
> The expected results would be that the third darcs either adds the
> replace change or complains about weird token-chars then exits with
> non-zero value. Previous replaces should not be lost.

Replace does not work if one of the tokens contains a space. Hm,
I thought this bug was fixed long ago, but maybe it's not. Darcs
would create a wrongly formated patch in pending when tokens
contain spaces, but I think it only made the replace show up as
hunks instead. Well, I'm familiar with this piece of code and
can take a look at it tomorrow, unless anyone beats me to it.  :-)

fixed in 1.0.9