From 5972418d66b3ae0e9c4ed76d5654d8681b59a436 Mon Sep 17 00:00:00 2001
From: "riseproject-dev[bot]"
 <330740410+riseproject-dev[bot]@users.noreply.github.com>
Date: Thu, 1 Oct 2026 08:33:02 +0000
Subject: [PATCH 1/2] djlint: Add versions 1.46.3, 1.46.4

Signed-off-by: riseproject-dev[bot] <330740410+riseproject-dev[bot]@users.noreply.github.com>
---
 docs/packages/djlint.yaml | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/docs/packages/djlint.yaml b/docs/packages/djlint.yaml
index 3d6fdf31b0f..4a830a18b14 100644
--- a/docs/packages/djlint.yaml
+++ b/docs/packages/djlint.yaml
@@ -107,3 +107,5 @@ versions:
   - filename: djlint-1.46.2-cp314-cp314t-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl
     sha256: f9ad85810ad93d3474042be0a03877554c813de9edc61f26c28b852f8a348f83
     requires-python: '>=3.10'
+- version: 1.46.3
+- version: 1.46.4

From 5c5306f813acdb5d1a4052f6473c07ada60902a4 Mon Sep 17 00:00:00 2001
From: Ludovic Henry <git@ludovic.dev>
Date: Thu, 1 Oct 2026 08:49:21 +0000
Subject: [PATCH 2/2] djlint: deselect test_lint[blocks], a hard-coded timing
 budget this runner can't meet

New in 1.46.3: the test asserts time.perf_counter() - started < 5 against
a pathological input of thousands of nested {% %} tags. Measured 8-24s
across all four interpreters we build (cp312/cp313/cp314/cp314t) -
deterministically over budget on this runner, not a flake.
---
 .github/workflows/build-djlint.yml | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/build-djlint.yml b/.github/workflows/build-djlint.yml
index 0b4cd647b55..e4a0cc6b89c 100644
--- a/.github/workflows/build-djlint.yml
+++ b/.github/workflows/build-djlint.yml
@@ -110,9 +110,17 @@ jobs:
           # settings.py").read_text()) to diff its option list against the
           # published JSON schema - same shape as rules.yaml above, just a
           # missing CIBW_TEST_SOURCES entry rather than a real test bug.
+          #
+          # test_lint[blocks] (new in 1.46.3) deselected: it asserts
+          # time.perf_counter() - started < 5 against a pathological input of
+          # thousands of nested {% %} tags. Measured 8-24s across our
+          # cp312/cp313/cp314/cp314t runners - consistently and deterministically
+          # over budget, not a flake. A hard-coded upstream performance
+          # assumption this riscv64 runner can't meet, same shape as
+          # test_closed_pipe_is_not_a_crash above.
           CIBW_TEST_COMMAND: >-
             python -c "import djlint; assert djlint.__file__.endswith('.so'), djlint.__file__"
-            && python -m pytest tests --deselect tests/test_djlint/test_djlint.py::test_closed_pipe_is_not_a_crash
+            && python -m pytest tests --deselect tests/test_djlint/test_djlint.py::test_closed_pipe_is_not_a_crash --deselect tests/test_unclosed_template_tags.py::test_lint[blocks]
 
       - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a  # v7.0.1
         with: