Sometimes patches with one "atomic change" have curly braces and sometimes they
don't. Specificaly, patches imported with 'darcs get --hashed' retain curly
braces, while patches newly created in the hashed repository are not represented
with curly braces.

See the following example. The first two patches were imported with 'darcs get
--hashed', while the second two patches were created directly in the hashed
repository.

[test setup
test@example.com**20080518153437] {
addfile ./baz
hunk ./baz 1
+test setup again
}

[patch format test
test@example.com**20080518153518] {
hunk ./baz 2
+darcs1 test one hunk
+
}

[local one hunk
test@example.com**20080518173216] hunk ./baz 3
+
+one hunk locally recorded in darcs2

[two hunk local patch
test@example.com**20080518173647] {
hunk ./baz 1
-test setup again
+test setup
hunk ./baz 4
-one hunk locally recorded in darcs2
+two hunk locally recorded in darcs2
}

Darcs 2.0.0 stable release binary for MacOS X. Patches were listed with 'darcs
send -o'.

Jason: any chance this could be serious?

Ok, I'm going to escalate because I can't tell if this is worrisome or not, and
I'd rather be safe than sorry.  David: is there any need to be worried about
this bug?

Do you consider this a bug? If so, why?

David

I guess the potential cause for concern was something about hashes being
different for otherwise the same patch created under the different circumstances
(sort of the opposite problem of a collision).  It didn't seem like we really
relied on this for anything in particular, but thought it was worth checking :-)

Otherwise, I think why Ben considers it to be a bug is because it makes it
harder for outside tools be parsing darcs patch files, definitely a lower
priority than the above.

Hi Ben, 

I think it's probably best that we left this alone.  If anybody wants to parse
darcs patches, they'll just have to account for the possibility that (in the
case of there only being a single primitive patch in the named patch) there may
not be any curly braces.  

My thinking is that any tool that works with darcs patches is going to have to
deal with old patch files produced by darcs anyway, so this spending time to fix
this probably won't really help matters.

Thanks!