8.3.2.38. v4.6.2ΒΆ

commit 45cdac2f70faf95ced53179c8bde450e1b82d99b
Author: Greg Althaus <galthaus@austin.rr.com>
Date:   Tue May 11 21:03:37 2021 -0500

    build: update to v4.6.2

M   go.mod
M   go.sum

commit fcd95526bfbc3e5f0270d7dbe28c89eafa6d0c3c
Author: Greg Althaus <galthaus@austin.rr.com>
Date:   Tue May 11 11:48:20 2021 -0500

    fix: username to allow @

M   go.mod
M   go.sum

commit 699ae9631042c7a609d43d69d5af56778df2df8f
Author: Greg Althaus <galthaus@austin.rr.com>
Date:   Tue May 11 09:39:31 2021 -0500

    fix(backend): Update other paramers to match machines

M   backend/endpoints.go
M   backend/plugins.go
M   backend/profiles.go
M   backend/raw_model.go
M   backend/stage.go

commit 208b76ec86eb2987d49b7b586835ef15f6989263
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Mon May 10 15:17:45 2021 -0500

    feat(printArgs): Print interesting startup options.

    All startup options (from any source) that are different from their
    defaults will be printed, along with the fully expanded paths of all
    the working directories dr-provision uses.

M   server/args.go

commit 8535359749c4a988e08ede48389941426f0182d0
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Thu May 6 12:53:28 2021 -0500

    fix(import): Fix several legacy data import bugs.

    The legacy data import code would delete the legacy files after all
    the data had been loaded into memory, but before we had finished
    saving it all in the new format.  Oops.  We now save everything before
    removing the legacy files.

    Importing into consensus was much slower than it needed to be due to
    the database operating in sync write mode.  While this is correct for
    normal operation, it is not required when doing the initial bulk
    import.  Refactor the initial data load code to have the consensus
    database run with async writes permitted during initial data import.

    Fix up argument processing to force BaseRoot to be an absolute path
    before fixing up the rest of the args.  While at it, switch to using
    filepath.IsAbs instead of the broken-on-Windows-anyways test we were
    using.

    Make startup profiling actually work again.

M   clitest/test-data/output/TestLogsCli/logs.get/stdout.expect
M   consensus/raft.go
M   consensus/raftStore.go
M   consensus/raft_test.go
M   datastack/stack.go
M   server/args.go

commit fb6a529f1547fc56133e4e2f446b83b5fef8f713
Author: Greg Althaus <galthaus@austin.rr.com>
Date:   Wed May 5 10:58:23 2021 -0500

    fix(render): make clones of objects to pass to template engine

M   backend/renderData.go
M   backend/renderData_test.go

commit b8d682ba6235b81651b27074dce11ec1557a1132
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Wed May 5 09:52:06 2021 -0500

    fix(race): Some more race detector placation.  Hopefully

M   consensus/raftFSM.go

commit 51db01b23de165c4f80cb7cb0cba3575b1e678d5
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Tue May 4 16:25:06 2021 -0500

    fix(etag): Fix race condition when messing with dominance.

    Derp, messed up racy handling of keeping track when the leader last
    issued dominance.  Fix that.

M   datastack/etags.go
M   go.mod

commit b321fe4d25e595ba1830facc5b25ba684fc6cb2c
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Tue May 4 14:08:38 2021 -0500

    fix(etag) More startup work avoidance.

    There were a few places where we were doing unneeded repeat work
    calculating SHA checksums and exploding isos when the bootenv was
    mising a SHA checksum.  Switch to relying on etags and using them to
    avoid more work.  This may result in a one-time reextraction of
    previously extracted isos due to having the canary files in the
    exploded trees rely on the etag saved shasum instead of the one from
    the bootenv.

M   backend/bootenv.go
M   datastack/etags.go
M   embedded/assets/explode_iso.sh

commit a0d6311dcb4167380798e22dfebc93f38e73df75
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Tue May 4 11:56:10 2021 -0500

    perf(etags): Allow more etag parallelism.

    Rework the locking in the etags manager to allow for more overall
    parallelism and reduce the amount of potential duplicate work when we
    need to actually calculate a sha256sum.

    While we are at it, swizzle up plugin provider loading to use etags
    instead of its own sha256 calculation routines.

M   backend/pluginProvider_test.go
M   backend/pluginProviders.go
M   datastack/etags.go
M   datastack/pluginProvider.go
M   datastack/stack.go
M   go.mod

commit 49d80cfd76bfd99133286a7c0958318fce1bf05e
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Mon May 3 10:57:05 2021 -0500

    perf(etags): Pass stat info around instead in etags.

    While we are at it, fix up ISO SHA checking to use etags instead
    of manually recalculating the SHA for an ISO and wasting everyone's
    time doing it.

M   backend/bootenv.go
M   datastack/etags.go

commit dd30ebac4d6a6c45bdf1d16b904191923a3103b3
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Wed Apr 28 15:41:31 2021 -0500

    feat(jobs): Server-side implementation of jobs purge.

    This API call will remove all jobs that refer to a nonexistent
    machine, along with all jobs that are considered purgable according to
    the current job log retention preferences.

M   backend/jobs.go
M   clitest/test-data/output/TestJobCli/jobs/stdout.expect
M   frontend/jobs.go
M   go.mod
M   go.sum

commit 4f9959866bf87fc28b9f69a9d942929e63af0fc5
Author: Greg Althaus <galthaus@austin.rr.com>
Date:   Tue May 4 12:59:15 2021 -0500

    build(stack): change the basicstore to v4.6.0 version

M   clitest/test-data/output/TestContentCli/contents.list.2/stdout.expect
M   clitest/test-data/output/TestContentCli/contents.list.3/stdout.expect
M   clitest/test-data/output/TestContentCli/contents.list.4/stdout.expect
M   clitest/test-data/output/TestContentCli/contents.list.5/stdout.expect
M   clitest/test-data/output/TestContentCli/contents.list/stdout.expect
M   clitest/test-data/output/TestContentsFunctionalCli/contents.list.2/stdout.expect
M   clitest/test-data/output/TestContentsFunctionalCli/contents.list/stdout.expect
M   clitest/test-data/output/TestPluginProviderCli/extended.4f8f2a612fb294982641072a65234ac0/stdout.expect
M   clitest/test-data/output/TestPluginProviderCli/extended.abe26401cab1af6c8ca1774b079184dd/stdout.expect
M   datastack/content.go

commit fb145be668b9c7da5ad3baa4d4253f79c2d9d6f9
Author: Greg Althaus <galthaus@austin.rr.com>
Date:   Tue May 4 11:16:42 2021 -0500

    fix(stack): allow raw_model ids to be sortable

M   datastack/stack.go

commit 692b203196cb4d19cc20ebaca594daec3141609b
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Mon May 3 13:14:08 2021 -0500

    fix(msgs): Tune a couple of error messages.

    Get rid of the sealing claims message, it was there for debugging.

    Make the secure bootlaoders messages audit instead of error, and make
    them slightly less scary.

M   backend/runningPlugin.go
M   midlayer/pxe.go

commit 4c880bb883f54ee5b875396a937fa018b575502d
Author: Victor Lowther <victor.lowther@gmail.com>
Date:   Mon May 3 11:39:19 2021 -0500

    perf(badger): Tune the value log sizes and GC.

    Also catch the error that a value log needs truncating and reopen with
    Truncate=true once.  If that does not allow the database to open, we
    will error out.

    Truncation needs to happen when there is garbage at the end of a value
    log, which can happen when a write was aborted halfway through due to
    running out of disk space or something.

M   consensus/raftStore.go
A   failover_test/vlog_test.go

End of Note