mirror of http://darcs.net/reviewed (http://darcs.net/Development/GettingStarted)

adjust filepath, directory and temporary deps to debian stretch

ghSat Mar 24 17:00:11 UTC 2018

resolve issue2579: darcs send will not work if no MTA is installed

ghFri Mar 23 21:40:10 UTC 2018

fix writing of hex strings in Darcs.Util.ByteString

Numbers a to f should we written lowercase, otherwise internal patch files and bundles become incompatible with older darcsen.

ghFri Mar 30 17:40:47 UTC 2018

rollback definition of linePS and use singleton where appropriate

ghThu Mar 29 22:46:05 UTC 2018

throw out all access to bytestring internals from Darcs.Util.ByteString

The only remaining function, unsafeWithInternals, is used only in Darcs.Util.Hash so it has been moved into that module. The two C files src/fpstring.[ch] have been removed.

bfrWed Mar 28 19:07:51 UTC 2018

fix betweenLinesPS and add a unit test

This is part of the program to get rid of using ByteString internals. The previous version of betweenLinesPS was buggy because it relied on implementation details of unlinesPS, see issue2573.

bfWed Mar 28 15:56:56 UTC 2018

re-format import list of Darcs.Util.ByteString

bfrWed Mar 28 15:47:39 UTC 2018

refactor, simplify, and document hunk application

This refactor does two things: First, it decouples the optimization that applies sequences of hunks to the same file in one go from the algorithm that applies a single hunk. Second, single hunk application is somewhat simplified and extensively documented. This is very subtle and highly optimized code that easily breaks.

This also moves breakAfter/BeforeNthNewline to D.P.Prim.V1.Commute. These functions are only used in this one place. They are undocumented and have subtle semantics, which makes them unsuitable for re-use.

bfThu Feb 22 09:58:00 UTC 2018

more efficient definition of unlinesPS

bfrThu Feb 22 19:14:56 UTC 2018

small simplification in D.R.Rebase

bfWed Mar 28 06:08:07 UTC 2018

fixed/extended haddocks in D.P.Rebase, consolidate error cases

bfTue Mar 27 18:22:46 UTC 2018

D.UI.Commands.Rebase: add a few comments to explain what's going on

bfTue Mar 27 18:15:56 UTC 2018

make spurious failures of hashed storage tests less likely

This is not a complete fix, as I am still seeing occasional failures ("harness/hstestdata.zip: openBinaryFile: does not exist").

bfrWed Mar 28 17:55:44 UTC 2018

allow running the diff.sh test in parallel

We use our own temporary directory for each test to avoid spurious differences between outputs of 'darcs diff' due to temporary names being already taken.

bfrWed Mar 28 17:18:40 UTC 2018

allow shell tests to run in parallel when using --myers and --patience

Like with --darcs1 and --darcs2, we build this option into the path where the test is run.

bfrWed Mar 28 17:17:07 UTC 2018

tests/network: bake process id into REMOTE_DIR

This allows to run the network tests in parallel.

bfSun Feb 25 23:42:39 UTC 2018

tests: fix add.sh that created repos directly in /tmp

bfSun Feb 25 23:41:09 UTC 2018

move makeFilename and readPatchInfos to D.R.Old

These functions are only used to read old-fashioned (non-hashed) repos. Concentrating them inside Darcs.Repository.Old (particularly makeFilename) makes it clear that they only exist for compatibility and not to be used elsewhere.

bfSun Oct 22 10:23:24 UTC 2017

cleanup: rename x to fmt in D.P.Prim.V1.Show

bfSun Feb 25 12:40:58 UTC 2018

remove several unused exports of D.R.Hashed

addToSpecificInventory is particularly dangerous, since it doesn't update the witnesses of the repo. Its only use is in addToTentativeInventory which is similarly dangerous and unfortunately is used (in convert import); it now has haddocks which warn about that.

bfTue Mar 27 18:25:21 UTC 2018

add newtypes to separate hashes for patches, inventories, and pristine trees

This makes type signatures more helpful and avoids errors when hashes to different kinds of objects are involved. Currently limited to D.R.Hashed and D.R.Inventory and defined in the latter, but it really belongs in D.R.Cache or a separate module.

This is an extension and generalization of the ValidHash type introduced previously. ValidHash is now a type class so we can use the same get/mkValidHash functions for the different hash types.

bfSat Mar 24 20:48:33 UTC 2018

fix, adapt, and add haddocks in D.R.Hashed

bfSun Mar 18 22:18:28 UTC 2018

cleanups in D.R.Inventory (layout, comments)

bfSat Mar 17 13:42:22 UTC 2018

pass the full path name of the head inventory to readInventoryPrivate/Local

bfTue Mar 13 13:02:11 UTC 2018

removed all -fno-warn-deprecations pragmas in the test suite

It seems these are no longer needed.

bfTue Mar 13 09:11:33 UTC 2018

add -fno-warn-orphans for the whole test suite

...and remove the individual OPTIONS_GHC pragmas with the same effect. For an executable,orphan instances are not a serious problem and we use them in quite a lot of places.

bfTue Mar 13 09:02:45 UTC 2018

use parser combinators intead of ad-hoc functions to parse inventories

bfMon Mar 19 08:18:25 UTC 2018

add properties and unit tests for D.R.Inventory

For the usual parse/show roundtrip property we need a newtype wrapper for Strings that are valid hashes according to D.R.Cache.okayHash; which is a good idea anyway, so we expose ValidHash as part of the Inventory type.

bfMon Mar 12 23:12:04 UTC 2018

add DirLayout data type to abstrcat over plain vs. bucketed in D.R.Hashed

bfSun Mar 11 11:38:55 UTC 2018

factor D.R.Hashed.Inventory out of D.R.Hashed, add data type for inventories

This is a first step to making the huge D.R.Hashed more modular.

bfSun Mar 11 10:38:15 UTC 2018