Edward Amsden
5336ca7a91
Merge pull request #201 from urbit/barter-simsum/file-extension
...
PMA file extension
2024-01-17 17:39:10 -06:00
barter-simsum
85a878a85c
pma: remove comments referring to the persistent file's "frontier"
2024-01-17 18:07:56 -05:00
Philip Quirk
25239fd475
don't fsync on every nock event while tracing
2024-01-16 12:04:42 -06:00
barter-simsum
390c1c3356
pma: file extension bug fixes. test changes
...
file extension seems to be working now after reordering freelist restoration and
calling _bt_falloc in _nlist_new to create the first partition
should be able to move on to partition striping now and the explicit call to
_bt_falloc will make this easier
2024-01-12 20:02:59 -05:00
barter-simsum
a604f19d74
pma: nlist creation depends on flist existence
...
reordered to handle this dependency. Still debugging an issue with _flist_grow
The dependency is natural and will be required to implement partition
striping. All node partitions should call falloc and be represented in the flist
2024-01-12 19:41:19 -05:00
barter-simsum
aaa6f5e2ff
pma: persistent file growth highpg tracking attempted fix
...
spent a while debugging this, but there are still issues on restoration. We
cannot simply call _flist_new and _flist_record_alloc on persistent state
restoration since _flist_new won't properly set the highpg. If restoring, we
should derive the high flist page (alloced or not) from state->file_size_p. I
think I way overcomplicated this
2024-01-12 18:11:42 -05:00
barter-simsum
6a08d08c0f
pma: file resize test in btest.c and fixes to resize logic
2024-01-11 18:26:33 -05:00
barter-simsum
f6213bd1ce
pma: minor cleanup
2024-01-10 19:15:34 -05:00
barter-simsum
aa21d14739
pma: remove frontier and only rely on state->file_size_p
2024-01-10 18:49:55 -05:00
barter-simsum
19fd0e4fb4
pma: file extension - to be revised
2024-01-10 18:27:31 -05:00
Edward Amsden
a96c669b22
Merge pull request #193 from urbit/jon/fix-rev
...
jets: fix jet_rev
2023-12-20 11:05:22 -06:00
barter-simsum
4c8851c385
pma: simpler pending freelist merge. fixes use after free
2023-12-20 09:37:04 -05:00
Edward Amsden
4080f50145
pma: format
2023-12-20 00:20:10 -06:00
Edward Amsden
3ae877f655
pma: lint and c warnings cleanup
2023-12-20 00:16:55 -06:00
Edward Amsden
58c49bc5e1
pma: dont assert loidx != in _bt_dirty, it could well be!
2023-12-19 21:36:27 -06:00
Edward Amsden
86a8922e0f
pma: make sure unifying equality dirties for unifying a cell in the PMA, not just an indirect atom
2023-12-19 20:56:33 -06:00
Edward Amsden
feeb278794
pma: lots of c-side bugfixes
2023-12-19 20:55:57 -06:00
Edward Amsden
2c1f1fa285
Merge branch 'status' into eamsden/integrate-pma-tmp
2023-12-19 18:38:40 -06:00
barter-simsum
2762872a94
pma: wip test revisions
2023-12-19 17:48:04 -05:00
barter-simsum
0dac274a63
pma: minor bug fixes
2023-12-19 17:47:57 -05:00
barter-simsum
232a5bdff5
pma: freelist restoration using *_record_alloc and insertdat bugfix
2023-12-19 16:29:03 -05:00
Jonathan Paprocki
16c91e96d9
jets: rev PR comment
2023-12-19 13:56:46 -05:00
Jonathan Paprocki
24ce412a1f
jets: fix jet_rev
2023-12-19 12:58:37 -05:00
Edward Amsden
f601a61827
pma: clean up a bunch of warnings (not all though)
2023-12-19 09:56:34 -06:00
barter-simsum
9a74aff3e8
pma: significant revisions to freelist logic
2023-12-18 22:02:58 -05:00
Philip Quirk
760b317a3e
add a warning to discourage use of check_all feature flag
2023-12-18 13:48:20 -06:00
barter-simsum
c0e68a27f6
pma: btest ephemeral structure equality tests
2023-12-17 20:05:03 -05:00
Alex Shelkovnykov
be96c495a8
serf: implement %live %exit
2023-12-17 14:00:19 -03:00
barter-simsum
3a6ed42054
pma: depth passed to _flist_read2 should be 1 not 0
2023-12-15 19:49:35 -05:00
barter-simsum
2c016500ac
pma: restore mmaps before regenerating ephemeral state
2023-12-15 19:24:37 -05:00
barter-simsum
9a8229837d
pma: initialize state->which to 0 in bt_state_open
2023-12-15 18:41:50 -05:00
barter-simsum
0ec984663c
pma: revisions to codepath handling opening of existing pma
2023-12-15 18:38:52 -05:00
barter-simsum
0496de5f4c
pma: testing out a mmap call /before/ reading header
2023-12-15 16:01:34 -05:00
Edward Amsden
012a7bac22
Merge pull request #180 from urbit/sigilante/remove-doublejam
...
serf: removal of snapshotting
2023-12-15 13:41:18 -06:00
Edward Amsden
011d452156
main: add stop_for_debug feature to stop Ares when it starts for a debugger to attach
2023-12-15 13:12:05 -06:00
Edward Amsden
820b1ac57a
main: add stop_for_debug feature to stop Ares when it starts for a debugger to attach
2023-12-15 12:27:17 -06:00
Edward Amsden
f092c482ee
pma: format
2023-12-15 11:17:08 -06:00
Edward Amsden
dd814b3e1f
pma: add crate feature to control debugging printfs
2023-12-15 11:16:38 -06:00
Sigilante
a4d67e01fd
serf: placate linter
2023-12-15 10:39:02 -06:00
Edward Amsden
3acf8e74c1
serf: ensure locals are preserved and top frame flipped after PMA save
2023-12-15 10:00:38 -06:00
barter-simsum
02b1ae3dd2
pma: btest.c changes
2023-12-15 06:36:53 -05:00
barter-simsum
66aaa255f7
pma: fix freelist node size calculation in _mlist_insert
2023-12-15 06:36:30 -05:00
barter-simsum
031e583c3c
pma: fix DPRINTF in bt_malloc
2023-12-15 05:38:35 -05:00
barter-simsum
965865da13
pma: mmap freespace ahead of first node partition
2023-12-15 03:24:42 -05:00
barter-simsum
40a04ec8ab
pma: _bt_printnode print to stderr
2023-12-15 01:19:28 -05:00
Edward Amsden
a37cf3e779
pma: when persisting hamt check if buffer already in PMA
2023-12-15 00:15:40 -06:00
barter-simsum
ff204263ff
pma: misc fixes
...
- primarily fixes a bug in _mlist_insert and _pending_flist_insert
- also updates some tests
- disables node data printing on _bt_insertdat with DEBUG_PRINTNODE macro
2023-12-15 00:50:26 -05:00
Edward Amsden
bdccac7ff2
pma: fixup of global static PMA
2023-12-14 22:17:25 -06:00
Edward Amsden
8fb0953247
pma: dirty in unifying equality
...
This changes the PMA to be a static global instead of a threaded-through handle.
We use a OnceLock to enforce one-time-only opening of the PMA. There is
no public PMA type any more, just a collection of module-level methods
which act through the global reference.
This is consistent with the assumptions of the underlying C code, that
there will be only one PMA open in a process, and avoids the need to
thread the PMA to unifying equality calls.
2023-12-14 21:42:31 -06:00
Edward Amsden
caeea6ec2d
unifying equality: move to own module
2023-12-14 18:08:55 -06:00
