Releases: urbit/vere
vere-v3.0-rc5
Description
During pre-release testing of vere-v3.0-rc4 the ship ~dinleb-rambep experienced loom corruption caused by the refcount of a jet registration noun being off by one. We were unable to reproduce the issue but decided to revert #538 out of an abundance of caution. We also added the --gc-early subcommand to recover the potentially existing ships that had been running release candidate with the issue.
What's Changed
- Revert "u3: restore old-style, efficient reap to u3m_love()" by @pkova in #612
- vere: adds --gc-early flag to pack and meld subcommands by @pkova in #615
Full Changelog: vere-v3.0-rc4...vere-v3.0-rc5
vere-v3.0-rc4
Description
During pre-release testing of vere-v3.0-rc3 a bug was discovered with the free list migration. This bug caused spurious "loom: corrupt" warnings for some ships when upgrading from vere-v2.12 to vere-v3.0. Another bug was discovered in error handling of the urbit chop command, causing problems if the ship crashed exactly during a chop.
This release candidate includes #602 and #608 to fix these bugs.
What's Changed
- vere: fix epoc lifecycle error handling, enforce invariants by @joemfb in #602
- http: do not double free scry cache path by @pkova in #606
- v3-migrate: ream before reclaiming bytecode caches by @pkova in #608
Full Changelog: vere-v3.0-rc3...vere-v3.0-rc4
vere-v3.0-rc3
Description
During pre-release testing of vere-v3.0-rc2 a bug was discovered with LMBD initialization where if your ship does not fit in a direct atom certain migration scenarios could crash. We also noticed a regression in the chop command. This release candidate fixes those issues in addition to some typos in ames czar printfs.
What's Changed
- ames: fix ames czar prints by @yosoyubik in #595
- epoc: fix log metadata write bug, make pre-release repair more robust by @joemfb in #596
- vere: fix chop by @joemfb in #599
Full Changelog: vere-v3.0-rc2...vere-v3.0-rc3
vere-v3.0-rc2
Description
During pre-release testing of vere-v3.0-rc1 a bug was discovered in situations where a ship would crash exactly when performing the migration to the epoch system. This scenario together with an attempt to downgrade back to vere-v2.12 would result in the pier being left in a broken state. #592 fixes this problem.
This release candidate also includes a fix for MDNS to distinguish between fake ships and real ships on the local network.
What's Changed
- mdns: distinguish fake ships in the mdns namespace by @pkova in #584
- bazel: bump zlib by @pkova in #590
- fix epoch migration bug by @matthew-levan in #592
Full Changelog: vere-v3.0-rc1...vere-v3.0-rc2
vere-v3.0-rc1
What's Changed
- jets: match hoon changes to bif:by by @ashelkovnykov in #439
- Merge develop to next/kelvin/411 by @pkova in #505
- pier: decrement arvo kelvin by @pkova in #506
- Merge master back to develop after vere-v2.12 by @pkova in #517
- Update VERSION to 3.0 by @pkova in #518
- Epoch System by @matthew-levan in #459
- u3: rewrites jam atom encoding to maximize its range by @joemfb in #363
- Fixed next/kelvin workflow to not upload to docker by @mopfel-winrux in #529
- vere: support stack traces on failing boot / full replay by @joemfb in #523
- Fix advisory types in dor/mor jets. by @sigilante in #534
- hoon: 138k by @joemfb in #526
- epoc: cleanup by @joemfb in #531
- u3: inline cell deconstruction by @joemfb in #536
- u3: declares %139 jets for +scot and +scow by @joemfb in #537
- ames: resolve galaxy ips using ipv4 only by @pkova in #543
- vere: refactors event log metadata reading to avoid the loom by @joemfb in #547
- disk: check for null pointer when writing through fak_o by @pkova in #548
- serf: adds new, stateful memory-pressure thresholds by @joemfb in #546
- UIP-0103: Persistent Nock Caching + Loom Migration Framework + Pointer Compression Migration Refactor by @matthew-levan in #508
- u3: adjust free-list sizes by @joemfb in #539
- serf: reclaim from persistent memoization cache by @joemfb in #549
- ames: handle %nail gift, lanes to vere cache by @pkova in #541
- u3: adds unconditional freelist migration as workaround by @joemfb in #551
- jets: refactors and optimizes tree-math jets by @joemfb in #397
- Fix typo in ames.c by @sigilante in #550
- u3: redeclare inline u3x_atom for linkability by @joemfb in #552
- jet
+matearm by @sidnym-ladrut in #512 - fixes +fitz jet mismatch by @mikolajpp in #444
- build: fetch
urcryptinstead of vendoring it by @matthew-levan in #524 - u3: removes free-list ream from v3 migration by @joemfb in #553
- build: install
autoconf-archivein docker workflow by @matthew-levan in #554 - build: fix docker workflow by @matthew-levan in #555
- u3: restore old-style, efficient reap to u3m_love() by @joemfb in #538
- u3: optimizes u3i_edit (nock opcode 10) by @joemfb in #362
- merge develop into next/kelvin/411 by @joemfb in #560
- ames: account for variable-size num fragments by @yosoyubik in #561
- lick: Shut bug by @mopfel-winrux in #563
- mdns: add mdns for local network routing by @pkova in #511
- stun: minimal working client/server by @yosoyubik in #545
- ames: clean up libuv timers by @pkova in #568
- jets: declare %138 separately by @joemfb in #567
- ames, http: declare gang correctly when scrying by @pkova in #569
- bazel: change dbus source since the official site is down by @pkova in #572
- bazel: looks like this makes the docker CI work by @pkova in #573
- Merge develop to next/kelvin/411 by @pkova in #571
- ames: independent galaxy dns resolution for stun by @yosoyubik in #574
- ames: fix galaxy domain check by @yosoyubik in #575
- Merge next/kelvin/411 to develop by @pkova in #576
New Contributors
- @sidnym-ladrut made their first contribution in #512
- @mikolajpp made their first contribution in #444
- @yosoyubik made their first contribution in #561
Full Changelog: vere-v2.12...vere-v3.0-rc1
vere-v2.12
Description
vere-v2.12 declares compatibility with the upcoming 412k urbit-os release. It includes support for the %lick vane, modifications to galaxy packet forwarding and a bugfix to the %khan driver.
What's Changed
%lickvane by @mopfel-winrux in #358- pier: propagate pace to mars by @pkova in #473
- ames: forward iff we're a galaxy by @matthew-levan in #493
- add -flto=thin to mac aarch64 build cflags for ez 10% gain by @fighet-parnet in #469
- compile with flto by @barter-simsum in #487
- conn: stop crashing on small receiving buffers by @matthew-levan in #494
next: no-op when pace is set toonceby @matthew-levan in #498- Update VERSION to 2.12 by @pkova in #501
- build: upgrade
zlibto v1.3 by @matthew-levan in #503 - flake: update by @midlyx-hatrys in #507
New Contributors
- @fighet-parnet made their first contribution in #469
Full Changelog: vere-v2.11...vere-v2.12
vere-v2.12-rc1
What's Changed
%lickvane by @mopfel-winrux in #358- pier: propagate pace to mars by @pkova in #473
- ames: forward iff we're a galaxy by @matthew-levan in #493
- add -flto=thin to mac aarch64 build cflags for ez 10% gain by @fighet-parnet in #469
- compile with flto by @barter-simsum in #487
- conn: stop crashing on small receiving buffers by @matthew-levan in #494
next: no-op when pace is set toonceby @matthew-levan in #498- build: upgrade
zlibto v1.3 by @matthew-levan in #503 - flake: update by @midlyx-hatrys in #507
New Contributors
- @fighet-parnet made their first contribution in #469
Full Changelog: vere-v2.11...vere-v2.12-rc1
vere-v2.11
Description
This release fixes a guard page bounds check that would usually present as a ship crashing with loom: bogus ward. This bug (present since vere-v2.7) also has a small chance of corrupting persistent state, so we encourage users to upgrade as soon as possible.
The release also fixes a bug with the meld command that was causing it to run twice for every invocation.
What's Changed
- u3: move the guard page earlier when making roads by @joemfb in #479
- ci: fix error check in version upload by @joemfb in #481
- meld: once is enough by @joemfb in #483
Full Changelog: vere-v2.10...vere-v2.11
vere-v2.11-rc1
vere-v2.10
Description
This release includes runtime fixes mainly aimed at OTA publishers. A publisher running vere-v2.10 and urbit-os-v2.142 (releasing in a few days) will be able to reliably deliver the initial OTA through remote scry instead of falling back to ames.
The release also includes significantly better long-running replay capabilities that were missing since vere-v1.14. The <your-ship>/.run play command now supports the --auto-meld command line flag to automatically meld a long memory-intensive replay. It also supports the --snap-at x command line flag to snapshot every x events.
What's Changed
- u3: enforce a max of one page in south.bin by @joemfb in #458
- Add jets for min/max. by @sigilante in #457
- play: dedicated event replay, revived by @joemfb in #420
- fine: Loosen path parsing and bump page size by @philipcmonk in #454
- build: switch
gmpsource url by @matthew-levan in #462 - u3: "toss" ephemeral pages after every event by @joemfb in #422
- vere: map ephemeral space to file in pier by @philipcmonk in #410
- vere: print errors when failing to open lock-file by @joemfb in #465
- jets: let +bex create atoms 256MB or more by @philipcmonk in #464
- u3: improve snapshot corruption size checks by @joemfb in #468
New Contributors
- @sigilante made their first contribution in #457
Full Changelog: vere-v2.9...vere-v2.10