[whaon]

• I have read the CONTRIBUTING.md guidelines.
• I have registered the PR changes.

Ⅰ. Describe what this PR did

DM rejects a new xa start with XAER_RMERR while an in-doubt (prepared) branch of the same gtrid exists on the instance, so two branches of one Seata global transaction cannot hit the same DM instance in XA mode.

For dbType=dm, qualify the XA gtrid with the branch id (gtrid = xid + "-" + branchId, formatId 9753) so every branch owns a distinct gtrid. The legacy format (formatId 9752, gtrid = xid) is kept byte-for-byte for every other database.

Ⅱ. Does this pull request fix one issue?

fix #8310

Ⅲ. Why don't you add test cases (unit test/integration test)?

Ⅳ. Describe how to verify it

Ⅴ. Special notes for reviews

[funky-eyes]

Please make sure the email address in your Git commits matches the one you used to submit the PR.

[whaon]

Please make sure the email address in your Git commits matches the one you used to submit the PR.

done

[funky-eyes]

Reviewed HEAD 88e3e16dafa8b4ba83ccedcf415bb337b95f6275 against merge base a9b03977be324c170741ebc37d797d5bfd159552.

Requesting changes for the two inline issues: valid long XIDs can fail after branch registration and remain unresolvable through the new encoder, and pre-upgrade DM prepared branches cannot be reconstructed with their original physical XA identity.

Please also provide a minimal real-DM reproduction or integration test for #8310. The reported stack fails in DmdbXAResource.end(), whereas the PR describes rejection in start(). That does not disprove the proposed cause, but the current byte-array tests do not establish it. Compare shared-gtrid/different-bqual branches against the new encoding on the same DM instance, record the failing XA call/error code and server/driver versions, and verify commit and rollback, including RM restart.

Validation: compiled the PR's actual Xid sources, executed all 8 existing methods in XAXidBuilderTest and XABranchXidTest with their JUnit assertions (all passed), reproduced the 62-byte-to-81-byte IPv6 failure, and checked that legacy prepare/new-DM phase-two produce different physical Xid tuples. No real DM database or full Maven suite was run. The multi-schema resourceId collision predates this PR and is not included as a regression finding.

[codecov]

Codecov Report

❌ Patch coverage is 86.04651% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.87%. Comparing base (cba73e1) to head (24a7509).

Files with missing lines	Patch %	Lines
...ache/seata/rm/datasource/xa/ConnectionProxyXA.java	77.77%	1 Missing and 3 partials ⚠️
...org/apache/seata/rm/datasource/xa/XABranchXid.java	90.00%	2 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##                2.x    #8312      +/-   ##
============================================
+ Coverage     79.84%   79.87%   +0.03%     
+ Complexity     1162     1160       -2     
============================================
  Files          1400     1400              
  Lines         53011    53046      +35     
  Branches       6349     6357       +8     
============================================
+ Hits          42325    42371      +46     
+ Misses         7729     7723       -6     
+ Partials       2957     2952       -5     

Files with missing lines	Coverage Δ
...ache/seata/rm/datasource/xa/ResourceManagerXA.java	89.04% <100.00%> (+0.15%)	⬆️
...rg/apache/seata/rm/datasource/xa/XAXidBuilder.java	100.00% <100.00%> (ø)
...org/apache/seata/rm/datasource/xa/XABranchXid.java	75.80% <90.00%> (+7.62%)	⬆️
...ache/seata/rm/datasource/xa/ConnectionProxyXA.java	89.49% <77.77%> (-0.80%)	⬇️

... and 11 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.