Hi,

darcs get --to-match 'hash ....' fails with the following message:

  darcs: _darcs/tentative_pristine: openBinaryFile: does not exist (No such file or directory)

as this test shows (I will send a patch introducing it as soon as I
get back the issue number):

#!/bin/sh

set -ev

## I would use the builtin !, but that has the wrong semantics.
not () { "$@" && exit 1 || :; }

rm -rf temp1
mkdir temp1
cd temp1
darcs init
echo first > a
darcs add a
darcs record -am 'first'
firsthash=`darcs changes --xml | grep 'hash=' | sed -e "s/.*hash='//" -e "s/'>//"`
echo second > b
darcs add b
darcs record -am 'second'

cd ..
rm -rf temp2
mkdir temp2
cd temp2
darcs init
echo darcs pull -v -a --match "hash $firsthash" ../temp1
darcs pull -v -a --match "hash $firsthash" ../temp1

cd ..
rm -rf temp3
echo darcs get -v --to-match "hash $firsthash" temp1 temp3
darcs get -v --to-match "hash $firsthash" temp1 temp3

thank you,
ciao, lele.
-- 
nickname: Lele Gaifax    | Quando vivrò di quello che ho pensato ieri
real: Emanuele Gaifas    | comincerò ad aver paura di chi mi copia.
lele@nautilus.homeip.net |                 -- Fortunato Depero, 1929.

I'm upgrading this ticket to 'urgent' because it's a darcs failure, and also
because it helpfully includes a test case to trigger the bug, which should make
it easier to work on.

I believe the problem is here:
in writePatchSet (called by patchSetToRepository), this code here:

    if format_has HashedInventory rf2
       then do HashedRepo.write_tentative_inventory c opts patchset
               HashedRepo.finalize_tentative_changes repo opts

Now write_tentative_inventory does just that: writes tentative_hashed_inventory,
but it does *not* touch anything about tentative_pristine. But
finalize_tentative_changes wants that file. (Note that patchSetToRepository is
called in the "--to-match" path of darcs get.)

(Ah, and the documentation in patchSetToRepository seems to be wrong about
calling error when working with HashedRepo's? It seems to be supposed to work.)

IIUIC, we first need to create an "empty" pristine and write that, because in
hashed repositories, pristine pointer is part of hashed_inventory, which we then
cannot write without one (that's why finish_tentative_changes fails). The call to
pristineFromWorking will fill in the so-far empty pristine then.

If there are no objections to this analysis, I'll post patches in a while (maybe
today, maybe tomorrow).

There's a patch on darcs-users@ pending review now.

Patch's been rejected because of violating some sort of unwritten rule about
module imports. I have no intention to refactor the surrounding code to appease
these rules just to get the fix through -> unassigning, kthxbai.

If your code is going  to work around bugs rather than fixing
them, yes, I will ask you to rewrite it.  I gave you a simple way to
tell when this is happening, and now you're going to leave in a huff?
If you don't care about the quality and maintainability of darcs'
code, then we're better off without you.  If you do care about darcs'
quality, you should learn to accept feedback to improve your code.

The following patch updated the status of issue885 to be resolved:

* resolve issue885: fix patchSetToRepository to work with hashed.