Repository navigation
tests/windows: correct the xxreg known-failure diagnosis - #9
Merged
Merged
Conversation
The xxreg/xreg "2-/8-bit field" failure was misattributed to an MSVC enum/width bug. The real cause is a stale in-source generated meta-flow.inc (left by a prior autotools build of the ovs submodule) shadowing the freshly generated gen/ copy: a quoted #include resolves the compiland's own dir first, so the stale 183-entry table compiled against today's 211-id enum, misindexing mf_fields[] and giving xreg/xxreg the wrong widths. A clean checkout (CI) is unaffected. Removing the stale .inc fixes it (verified via `ovntest test-ovn parse-expr`); a configure-time guard in dbosoft/ovs CMake prevents recurrence and reaches OVN on the next ovs submodule bump. The local baseline was contaminated by this, so it must be re-measured on a clean build.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
known-failures.mdattributed thexxreg/xreg"2-/8-bit field" failure to an MSVC enum/width miscomputation. That diagnosis was wrong.Real root cause: a prior in-source autotools build of the
ovssubmodule left a stale generatedovs/lib/meta-flow.incon disk. A quoted#include "meta-flow.inc"fromovs/lib/meta-flow.cresolves the compiland's own directory before any-Ipath, so the stale copy shadowed the freshly generatedgen/lib/meta-flow.inc. The stale table had 183 field entries vs today's 211-idMFF_N_IDS; with positional init + direct indexing (mf_fields[id]), the register fields read the wrong slots and reported 2-/8-bit instead of 128/64-bit. OVN'slib/logical-fields.cwas correct all along.A clean checkout (CI) never had the stale
.inc, so CI was unaffected — this only bites a local tree that was once autotools-built. Confirmed fixed after removing it:ovntest test-ovn parse-exprnormalizesxxreg0[0..127]→xxreg0and acceptsxxreg0[64..127]with no error. A configure-time guard in dbosoft/ovs (PR dbosoft/ovs#51) prevents recurrence and reaches OVN on the nextovssubmodule bump.Doc-only: corrects the section-1 root cause, flags that the local 304/65 baseline was contaminated and must be re-measured on a clean build, and notes the ~37 cascades are expected to clear. Bug #2 (mirror Index/Key) is untouched.