This project is mirrored from https://github.com/cockroachdb/cockroach. Pull mirroring updated .
  1. 05 Aug, 2020 17 commits
    • craig[bot]'s avatar
      Merge #52381 · 42c8989c
      craig[bot] authored
      52381: sql: improve overload selection when multiple candidates require the same type r=RaduBerinde a=RaduBerinde
      
      Our type checking cannot deal with expression `('{"x": "bar"}') -> 'x'`.
      This works without the parens because literal constants are treated as
      a special case. It used to work with parens because of some constant
      folding that was done before type-checking (which was removed).
      
      A principled fix would require big changes in type-checking. However,
      I did notice that some of the logic for filtering overloads is too
      restrictive. We only try to `TypeCheck` binary op arguments with a
      specific type when we have a single candidate overload left. This is
      why we don't try to type-check `('{"x": "bar"}')` as a Jsonb.
      
      This change improves this heuristic: instead of checking whether we
      have a single remaining candidate, check if all remaining candidates
      require the same type for that argument. This is a net improvement
      that happens to help in this particular case (where we have two
      variants for `->`, both taking a Jsonb on the left).
      
      Fixes #52043.
      
      Release note: None
      Co-authored-by: default avatarRadu Berinde <[email protected]>
      42c8989c
    • craig[bot]'s avatar
      Merge #51803 #52083 #52094 #52352 #52386 · 67a92cd0
      craig[bot] authored
      51803: cmd/docgen: add HTTP extractor r=mjibson a=mjibson
      
      Add a way to extract docs from the status.proto HTTP endpoint. These
      can be imported into the docs project as needed.
      
      Release note: None
      
      52083: roachtest: small misc r=andreimatei a=andreimatei
      
      See individual commits.
      
      52094: rpc: implement tenant access control policies at KV RPC boundary r=nvanbenschoten a=nvanbenschoten
      
      Fixes #47898.
      
      Rebased on #51503 and #52034. Ignore all but the last 3 commits.
      
      This commit adds a collection of access control policies for the newly exposed tenant RPC server. These authorization policies ensure that an authenticated tenant is only able to access keys within its keyspace and that no tenant is able to access data from another tenant's keyspace through the tenant RPC server. This is a major step in providing crypto-backed logical isolation between tenants in a multi-tenant cluster.
      
      The existing auth mechanism is retained on the standard RPC server, which means that the system tenant is still able to access any key in the system.
      
      52352: sql/pgwire: add regression test for varchar OIDs in RowDescription r=jordanlewis a=rafiss
      
      See issue #51360. The bug described in it was fixed somewhat
      accidentally, so this test will verify that we don't regress again.
      
      Release note: None
      
      52386: opt: add SerializingProject exec primitive r=RaduBerinde a=RaduBerinde
      
      The top-level projection of a query has a special property - it can project away
      columns that we want an ordering on (e.g. `SELECT a FROM t ORDER BY b`).
      
      The distsql physical planner was designed to tolerate such cases, as they were
      much more common with the heuristic planner. But the new distsql exec factory
      does not; it currently relies on a hack: it detects this case by checking if the
      required output ordering is `nil`. This is fragile and doesn't work in all
      cases.
      
      This change adds a `SerializingProject` primitive which is like a SimpleProject
      but it forces serialization of all parallel streams into one. The new primitive
      is used to enforce the final query presentation. We only need to pass column
      names for the presentation, so we remove `RenameColumns` and remove the column
      names argument from `SimpleProject` (simplifying some execbuilder code).
      
      We also fix a bug in `ConstructSimpleProject` where we weren't taking the
      `PlanToStreamColMap` into account when building the projection.
      
      Release note: None
      Co-authored-by: default avatarMatt Jibson <[email protected]>
      Co-authored-by: default avatarAndrei Matei <[email protected]>
      Co-authored-by: default avatarNathan VanBenschoten <[email protected]>
      Co-authored-by: default avatarRafi Shamim <[email protected]>
      Co-authored-by: default avatarRadu Berinde <[email protected]>
      67a92cd0
    • Radu Berinde's avatar
      opt: add SerializingProject exec primitive · 509b76dd
      Radu Berinde authored
      The top-level projection of a query has a special property - it can project away
      columns that we want an ordering on (e.g. `SELECT a FROM t ORDER BY b`).
      
      The distsql physical planner was designed to tolerate such cases, as they were
      much more common with the heuristic planner. But the new distsql exec factory
      does not; it currently relies on a hack: it detects this case by checking if the
      required output ordering is `nil`. This is fragile and doesn't work in all
      cases.
      
      This change adds a `SerializingProject` primitive which is like a SimpleProject
      but it forces serialization of all parallel streams into one. The new primitive
      is used to enforce the final query presentation. We only need to pass column
      names for the presentation, so we remove `RenameColumns` and remove the column
      names argument from `SimpleProject` (simplifying some execbuilder code).
      
      We also fix a bug in `ConstructSimpleProject` where we weren't taking the
      `PlanToStreamColMap` into account when building the projection.
      
      Release note: None
      509b76dd
    • Andrei Matei's avatar
      ts: better error message for bad query · f38ebe1a
      Andrei Matei authored
      Release note: None
      f38ebe1a
    • Andrei Matei's avatar
      roachtest: adjust duration for local test · fb196f7a
      Andrei Matei authored
      A test had a non-sensically short duration when running locally.
      Actually running this test with such a short duration causes it to fail
      cause the metrics query that it wants to do errors out because the
      duration is too small in comparison with the downsampling period.
      
      Release note: None
      fb196f7a
    • craig[bot]'s avatar
      Merge #52380 · 1cbbf7d7
      craig[bot] authored
      52380: geos: also look at paths parenting os.Argv[0] r=ajwerner a=otan
      
      For cases like roachprod, the cockroach binary may be invoked from an
      alternate directory. This may fail to find the appropriate GEOS library.
      As such, also take into account os.Argv[0] when looking at parenting
      directories.
      
      Release note: None
      Co-authored-by: default avatarOliver Tan <[email protected]>
      1cbbf7d7
    • craig[bot]'s avatar
      Merge #52221 · f98fe0de
      craig[bot] authored
      52221: cli: make fewer assumptions about the origin of `stderr` in tests r=irfansharif a=knz
      
      This further removes a dependency on the knowledge that the `log`
      package knows anything about stderr, in CLI tests.
      
      Found this while working on #52200.
      
      Release note: None
      Co-authored-by: default avatarRaphael 'kena' Poss <[email protected]>
      f98fe0de
    • Oliver Tan's avatar
      geos: also look at paths parenting os.Argv[0] · 3f114589
      Oliver Tan authored
      For cases like roachprod, the cockroach binary may be invoked from an
      alternate directory. This may fail to find the appropriate GEOS library.
      As such, also take into account os.Argv[0] when looking at parenting
      directories.
      
      Release note: None
      3f114589
    • craig[bot]'s avatar
      Merge #51910 · 8f7a5967
      craig[bot] authored
      51910: ui: use StatementsPage from admin-ui-components r=koorosh a=koorosh
      
      Depends on https://github.com/cockroachdb/admin-ui-components/pull/9
      Depends on https://github.com/cockroachdb/yarn-vendored/pull/27
      Depends on https://github.com/cockroachdb/yarn-vendored/pull/28
      
      `StatementsPage` component is replaced with previously extracted version of StatementsPage in `admin-ui-components`.
      The new version (extracted) of `StatementsPage` has several changes that required to make the following adjustments:
      - `ActivateStatementDiagnostics` modal isn't connected components anymore so dispatching actions are pulled up to connected StatementsPage wrapper.
      - Previously, analytics functions (which track user interaction) were
      called directly from React components that made this functionality tightly coupled.
      Now, sagas used to watch actions related to user interaction and then call async logic in sagas.
      -- components expose only callback functions which are related to the
      the logic component is responsible;
      -- connected component is responsible to dispatch actions with proper
      payload to be properly then handled by sagas.
      Co-authored-by: default avatarAndrii Vorobiov <[email protected]>
      8f7a5967
    • craig[bot]'s avatar
      Merge #51483 · 37c10c43
      craig[bot] authored
      51483: geo/geomfn: implement ST_Translate r=otan a=vxio
      
      Implement ST_Translate on arguments {geometry,float8,float8}, which should adopt PostGIS behaviour.
      
      Release note (sql change): Implemented geometry builtin `ST_Translate`
      Co-authored-by: default avatarVincent Xiao <[email protected]>
      37c10c43
    • craig[bot]'s avatar
      Merge #52272 #52389 · 4f05bc81
      craig[bot] authored
      52272: backupccl: create type schema change jobs for restored types r=pbardea a=rohany
      
      Fixes #50322.
      
      This commit teaches restore to create type schema change jobs for
      restored type descriptors that were backed up in the middle of a type
      schema change.
      
      Release note: None
      
      52389: backupccl: fix bug when scattering to nodes with double digit nodeIDs r=pbardea a=pbardea
      
      Previously a RESTORE on a 10 node cluster would fail with an error
      indicating that the spans provided to the SplitAndScatter's processor
      were out of order.
      
      The SplitAndScatter processor output router expects a sorted list of
      spans, each represening a particular node in the cluster. Previously,
      the span for a given node would be too wide and might cover spans
      generated for other nodes. For example, the span for node 1 was
      previously, `node{1-2}`. However this meant that this span would contain
      the span for node 10. This change ensures that each span consists of the
      start key until the next key, rather than ending at the start key's
      `PrefixEnd`.
      
      Fixes #52387.
      
      Release note: None
      Co-authored-by: default avatarRohan Yadav <[email protected]>
      Co-authored-by: default avatarPaul Bardea <[email protected]>
      4f05bc81
    • Paul Bardea's avatar
      backupccl: fix bug when scattering to nodes with double digit nodeIDs · a367c7f5
      Paul Bardea authored
      Previously a RESTORE on a 10 node cluster would fail with an error
      indicating that the spans provided to the SplitAndScatter's processor
      were out of order.
      
      The SplitAndScatter processor output router expects a sorted list of
      spans, each represening a particular node in the cluster. Previously,
      the span for a given node would be too wide and might cover spans
      generated for other nodes. For example, the span for node 1 was
      previously, `node{1-2}`. However this meant that this span would contain
      the span for node 10. This change ensures that each span consists of the
      start key until the next key, rather than ending at the start key's
      `PrefixEnd`.
      
      Release note: None
      a367c7f5
    • Vincent Xiao's avatar
      geo/geomfn: implement ST_Translate · 9c753196
      Vincent Xiao authored
      Implement ST_Translate on arguments {geometry,float8,float8}, which
      should adopt PostGIS behaviour.
      
      Release note (sql change): Implemented geometry builtin `ST_Translate`
      9c753196
    • craig[bot]'s avatar
      Merge #51250 · 92ca16fb
      craig[bot] authored
      51250: colexec: switch tests from using subtests to logging the info r=yuzefovich a=yuzefovich
      
      This change reduces the output size with `-test.v` option from 23k to
      500, but there is no visible change in the running time of the whole
      package - it still takes about 20s with disabled batch size
      randomization.
      
      Fixes: #50501.
      
      Release note: None
      Co-authored-by: default avatarYahor Yuzefovich <[email protected]>
      92ca16fb
    • craig[bot]'s avatar
      Merge #51362 · d5006156
      craig[bot] authored
      51362: sql: drop types as part of a `DROP DATABASE` statement r=rohany a=rohany
      
      Fixes #49940.
      
      This commit adds logic to drop all types as part of a `DROP DATABASE`
      cascade.
      
      Release note: None
      Co-authored-by: default avatarRohan Yadav <[email protected]>
      d5006156
    • craig[bot]'s avatar
      Merge #52261 #52319 #52372 · 3291543a
      craig[bot] authored
      52261: geosegmentize: prevent OOMs r=sumeerbhola a=otan
      
      geosegmentize can be given inputs that break a line down into too many
      coordinates. Guard against this by ensuring a maximum amount of
      coordinates allowed between points.
      
      PostgreSQL errors the allocate here too.
      
      Resolves #52218.
      Resolves #52217.
      
      Release note: None
      
      52319: roachtest: use simple query method for tpcc chaos tests r=RaduBerinde a=RaduBerinde
      
      By default, the TPCC workload uses prepared statements (for a subset
      of statements). All these statements are prepared on all connections
      upfront, which can take a while. If nodes "fail" (due to chaos), this
      process is restarted, which can happen over and over.
      
      To avoid this issue, we pass `--method=simple` for chaos tpcc and
      tpccbench tests. This means that the "simple" query protocol is used,
      where all arguments are rendered inside the query string.
      
      Fixes #52262.
      
      Release note: None
      
      52372: acceptance: disable Elixir tests r=rohany a=otan
      
      Elixir requires network to install dependencies. This can be changed
      (painfully), but an upgrade an Elixir is causing even more problems. I'm
      in favour of removing the acceptance (which I wrote) to fix this for now
      as I have plentiful unit tests covering the same thing. Maybe someone
      can come back to this if it's a large issue...
      
      Release note: None
      Co-authored-by: default avatarOliver Tan <[email protected]>
      Co-authored-by: default avatarRadu Berinde <[email protected]>
      3291543a
    • Radu Berinde's avatar
      sql: improve overload selection when multiple candidates require the same type · 1a9f0b5b
      Radu Berinde authored
      Our type checking cannot deal with expression `('{"x": "bar"}') -> 'x'`.
      This works without the parens because literal constants are treated as
      a special case. It used to work with parens because of some constant
      folding that was done before type-checking (which was removed).
      
      A principled fix would require big changes in type-checking. However,
      I did notice that some of the logic for filtering overloads is too
      restrictive. We only try to `TypeCheck` binary op arguments with a
      specific type when we have a single candidate overload left. This is
      why we don't try to type-check `('{"x": "bar"}')` as a Jsonb.
      
      This change improves this heuristic: instead of checking whether we
      have a single remaining candidate, check if all remaining candidates
      require the same type for that argument. This is a net improvement
      that happens to help in this particular case (where we have two
      variants for `->`, both taking a Jsonb on the left).
      
      Fixes #52043.
      
      Release note: None
      1a9f0b5b
  2. 04 Aug, 2020 23 commits
    • Oliver Tan's avatar
      acceptance: disable Elixir tests · 1dd102ac
      Oliver Tan authored
      Elixir requires network to install dependencies. This can be changed
      (painfully), but an upgrade on Elixir is causing even more problems. I'm
      in favour of removing the acceptance (which I wrote) to fix this for now
      as I have plentiful unit tests covering the same thing. Maybe someone
      can come back to this if it's a large issue...
      
      Release note: None
      1dd102ac
    • Rafi Shamim's avatar
      sql/pgwire: add regression test for varchar OIDs in RowDescription · b9600951
      Rafi Shamim authored
      See issue #51360. The bug described in it was fixed somewhat
      accidentally, do this test will verify that we don't regress again.
      
      Release note: None
      b9600951
    • Matt Jibson's avatar
      cmd/docgen: add HTTP extractor · 597f1ac5
      Matt Jibson authored
      Add a way to extract docs from the status.proto HTTP endpoint. These
      can be imported into the docs project as needed.
      
      Release note: None
      597f1ac5
    • Rohan Yadav's avatar
      backupccl: create type schema change jobs for restored types · 46f89bcb
      Rohan Yadav authored
      Fixes #50322.
      
      This commit teaches restore to create type schema change jobs for
      restored type descriptors that were backed up in the middle of a type
      schema change.
      
      Release note: None
      46f89bcb
    • Nathan VanBenschoten's avatar
      rpc: restructure requireSuperUser, avoid possible bug · 44983c4c
      Nathan VanBenschoten authored
      It's unclear whether this was exploitable. It probably wasn't because we
      should have already insisted on TLSInfo being present, but it can't hurt
      to restructure the code to prevent this kind of bug.
      44983c4c
    • Nathan VanBenschoten's avatar
      rpc: implement tenant access control policies at KV RPC boundary · f21b856d
      Nathan VanBenschoten authored
      Fixes #47898.
      
      Rebased on #51503 and #52034. Ignore all but the last 3 commits.
      
      This commit adds a collection of access control policies for the newly
      exposed tenant RPC server. These authorization policies ensure that an
      authenticated tenant is only able to access keys within its keyspace and
      that no tenant is able to access data from another tenant's keyspace
      through the tenant RPC server. This is a major step in providing
      crypto-backed logical isolation between tenants in a multi-tenant
      cluster.
      
      The existing auth mechanism is retained on the standard RPC server,
      which means that the system tenant is still able to access any key in
      the system.
      f21b856d
    • Nathan VanBenschoten's avatar
      keys: create SpanAddr utility · acfa3d47
      Nathan VanBenschoten authored
      This was badly needed in a number of places. I found myself reaching for
      it again, so I figured it was time to add it.
      acfa3d47
    • Nathan VanBenschoten's avatar
      sql: tenants don't unsplit manually split ranges on DROP {TABLE,INDEX} · d0fde99b
      Nathan VanBenschoten authored
      This requires us to scan the meta ranges, which is not something that
      secondary tenants are allowed to do. Leaving the ranges split until
      their sticky bit expires is fine, the existing code is just an
      optimization.
      
      This does serve as a reminder that:
      1. we should make sure to remove sticky bits when we destroy a tenant's
         keyspace (see #48775).
      2. we should probably not allow tenants to run `ALTER TABLE ... SPLIT AT`
         statements because we're not placing hard range count limits on tenants
         anywhere else. If others agree, I'll file an issue.
      d0fde99b
    • craig[bot]'s avatar
      Merge #52079 · 7ca4a462
      craig[bot] authored
      52079: cli: support user defined schemas in cockroach dump r=rohany a=rohany
      
      Fixes #50874.
      
      Release note (cli change): `cockroach dump` now supports dumping
      databases that contain user defined schemas, as well as accepting schema
      qualified table names as arguments.
      Co-authored-by: default avatarRohan Yadav <[email protected]>
      7ca4a462
    • craig[bot]'s avatar
      Merge #52215 · d054bba1
      craig[bot] authored
      52215: opt: modify JoinOrderBuilder lookup tables to handle special cases r=DrewKimball a=DrewKimball
      
      Some join transformations are only conditionally valid, depending
      upon the null-rejecting properties of the join predicates. For
      example, this transformation is only valid because both join
      predicates reject nulls on relation xy:
      ```
      SELECT * FROM
      (
        SELECT * FROM xy
        FULL JOIN ab ON x = a
      )
      FULL JOIN cd ON c = x
      =>
      SELECT * FROM
      (
        SELECT * FROM xy
        FULL JOIN cd ON c = x
      )
      FULL JOIN ab ON x = a
      ```
      This patch extends the join property lookup tables of
      join_order_builder.go to allow reordering in these cases.
      
      Release note: None
      Co-authored-by: default avatarDrew Kimball <[email protected]>
      d054bba1
    • Rohan Yadav's avatar
      cli: support user defined schemas in cockroach dump · 0de12f9c
      Rohan Yadav authored
      Fixes #50874.
      
      Release note (cli change): `cockroach dump` now supports dumping
      databases that contain user defined schemas, as well as accepting schema
      qualified table names as arguments.
      0de12f9c
    • craig[bot]'s avatar
      Merge #52204 · a5e5511e
      craig[bot] authored
      52204: sql: add PartialIndexMutator for generating random partial indexes r=mjibson a=mgartner
      
      This commit adds `PartialIndexMutator` for generating random partial
      indexes for randomized tests. The mutator is enabled for sqlsmith in
      this commit. Once the temporary `experimential_partial_indexes` session
      setting is removed, the mutation should be run for every invocation of
      `RandCreateTable`, similar to `IndexStoringMutator`.
      
      Fixes #52030
      
      Release note: None
      Co-authored-by: default avatarMarcus Gartner <[email protected]>
      a5e5511e
    • craig[bot]'s avatar
      Merge #52032 #52214 · 29083d75
      craig[bot] authored
      52032: sql: crdb_version & version() are the same r=rafiss a=ekalinin
      
      Fixes #51655
      
      Before this change License (info.Distribution) was not the same returned
      by "show crdb_version" and "select version();".
      
      Added function sql.makeReadOnlyVarWithFn to init "crdb_version" read
      only variables after "init" block.
      
      Release note (bug fix): crdb_version & version() return the same result
      
      52214: opt: panic when optbuilder encounters non-immutable partial index predicates r=RaduBerinde a=mgartner
      
      This commit adds a panic when optbuilder encounters a non-immutable
      partial index predicate expression while building select and mutation
      opt trees. Partials index predicates must be immutable, which is
      validated when they are created by `schemaexpr.IndexPredicateValidator`.
      If optbuilder encounters a predicate with non-immutable operators, a
      panic is preferred to producing incorrect query results.
      
      Release note: None
      Co-authored-by: default avatarEugene Kalinin <[email protected]>
      Co-authored-by: default avatarMarcus Gartner <[email protected]>
      29083d75
    • Marcus Gartner's avatar
      sql: add PartialIndexMutator for generating random partial indexes · 81a3faba
      Marcus Gartner authored
      This commit adds `PartialIndexMutator` for generating random partial
      indexes for randomized tests. The mutator is enabled for sqlsmith in
      this commit. Once the temporary `experimential_partial_indexes` session
      setting is removed, the mutation should be run for every invocation of
      `RandCreateTable`, similar to `IndexStoringMutator`.
      
      Release note: None
      81a3faba
    • craig[bot]'s avatar
      Merge #51305 #51882 #52146 #52320 · b380060e
      craig[bot] authored
      51305: geo,geoindex: use bounding box for geography coverings that are bad r=sumeerbhola a=sumeerbhola
      
      This change uses a covererInterface implementation for geography
      that notices when a covering is using top-level cells of all faces
      and in that case uses the bounding box to compute the covering.
      
      Also changed the bounding box calculation for geography shapes to
      use only the points and not first construct s2.Regions. The latter
      causes marginally bad shapes to continue to have bad coverings since
      the bounding box also covers the whole earth.
      
      Release note: None
      
      51882: roachpb: panic when comparing a Lease to a non-lease r=andreimatei a=andreimatei
      
      Release note: None
      
      52146: sql: remove local execution of projectSetNode and implement ConstructProjectSet in the new factory r=yuzefovich a=yuzefovich
      
      Depends on #52108.
      
      **sql: remove local execution of projectSetNode**
      
      We have project set processor which is always planned for
      `projectSetNode`, so this commit removes the dead code of its local
      execution. Additionally, it removes some unused fields and cleans up
      cancellation check of the processor.
      
      Release note: None
      
      **sql: implement ConstructProjectSet in the new factory**
      
      Addresses: #47473.
      
      Release note: None
      
      52320: kvserver: enable merges in kvnemesis r=aayushshah15 a=aayushshah15
      
      We had merges disabled because of the bugs tracked in #44878, but those have
      since been fixed by #46085 and #50265.
      
      Release note: None
      Co-authored-by: default avatarsumeerbhola <[email protected]>
      Co-authored-by: default avatarAndrei Matei <[email protected]>
      Co-authored-by: default avatarYahor Yuzefovich <[email protected]>
      Co-authored-by: default avatarAayush Shah <[email protected]>
      b380060e
    • Oliver Tan's avatar
      geosegmentize: prevent OOMs · b0d12625
      Oliver Tan authored
      geosegmentize can be given inputs that break a line down into too many
      coordinates. Guard against this by ensuring a maximum amount of
      coordinates allowed between points.
      
      PostgreSQL errors the allocate here too.
      
      Release note: None
      b0d12625
    • craig[bot]'s avatar
      Merge #51959 #52269 #52273 #52343 · 986e308b
      craig[bot] authored
      51959: backupccl: distribute RESTORE work using DistSQL r=dt,yuzefovich a=pbardea
      
      This commit replaces restore's old method of coordinating work on a
      single node to instead use DistSQL. It creates a 2 stage DistSQL flow.
      The first stage assigns chunks to SplitAndScatter processors which
      forward spans that they scattered to the second stage made up of the
      RestoreData processors which ingest the data. The SplitAndScatter
      processors attempt to send the work to the RestoreData colocated with
      the leaseholder of the span after the scattering.
      
      Release note: None.
      
      52269: roachtest: correctly start crdb in acceptance/rapid_restart r=andreimatei a=andreimatei
      
      --start-single-node was needed. Without it, I think the test's killing
      of a node raced with that node dieing by itself, and sometimes the race
      resulted in `cockroach stop` first seeing the process but then
      `/bin/bash: line 8: kill: (16016) - No such proces`
      
      Fixes #52060
      
      Release note: None
      
      52273: roachtest: remove confusing "--- PASS" log lines r=andreimatei a=andreimatei
      
      Two workloads were printing PASS/FAIL lines from inside the predicates
      passed to Searcher.Search. This is really confusing when reading the
      test's output, because they don't correspond to the test's disposition.
      
      Release note: None
      
      52343: roachtest: reflake (skip) disk-stalled test for release-19.1 r=irfansharif a=irfansharif
      
      Fixes #52181. This test doesn't make use of all the `roachtest start`
      smartness to figure out what sub-command to use and instead constructs
      the raw start command by hand.
      
      Given our support policy, let's just bump the minimum version this test
      expects to run. This feels like another instance where #51897 would be 
      nice to have.
      
      Release note: None
      Co-authored-by: default avatarPaul Bardea <[email protected]>
      Co-authored-by: default avatarAndrei Matei <[email protected]>
      Co-authored-by: default avatarirfan sharif <[email protected]>
      986e308b
    • sumeerbhola's avatar
      geo,geoindex: use bounding box for geography coverings that are bad · ddff188d
      sumeerbhola authored
      This change uses a covererInterface implementation for geography
      that notices when a covering is using top-level cells of all faces
      and in that case uses the bounding box to compute the covering.
      
      Also changed the bounding box calculation for geography shapes to
      use only the points and lines that are part of the shape instead
      of constructing s2.Regions. The latter causes bad shapes to
      continue to have bad coverings since the bounding box also covers
      the whole earth.
      
      Release note: None
      ddff188d
    • craig[bot]'s avatar
      Merge #52187 · 1d12a54a
      craig[bot] authored
      52187: build: add libgeos to teamcity-make-and-publish-build.sh r=jlinder a=otan
      
      Following the procedure outlined in bless.
      
      Release note: None
      Co-authored-by: default avatarOliver Tan <[email protected]>
      1d12a54a
    • Oliver Tan's avatar
      build: add libgeos to teamcity-make-and-publish-build.sh · b34acfeb
      Oliver Tan authored
      Following the procedure outlined in bless.
      
      Release note: None
      b34acfeb
    • Yahor Yuzefovich's avatar
      sql: implement ConstructProjectSet in the new factory · fe441898
      Yahor Yuzefovich authored
      Release note: None
      fe441898
    • Yahor Yuzefovich's avatar
      sql: remove local execution of projectSetNode · 8e463131
      Yahor Yuzefovich authored
      We have project set processor which is always planned for
      `projectSetNode`, so this commit removes the dead code of its local
      execution. Additionally, it removes some unused fields and cleans up
      cancellation check of the processor.
      
      Release note: None
      8e463131
    • irfan sharif's avatar
      roachtest: reflake (skip) disk-stalled test for release-19.1 · 41891988
      irfan sharif authored
      Fixes #52181. This test doesn't make use of all the `roachtest start`
      smartness to figure out what sub-command to use and instead constructs
      the raw start command by hand.
      
      Given our support policy, let's just bump the minimum version this test
      expects to run. This feels like another instance where
      https://github.com/cockroachdb/cockroach/issues/51897 would be nice to
      have.
      
      Release note: None
      41891988