Skip to content

test_unicodedata: test_normalization() fails randomly with IncompleteRead on PPC64LE Fedora buildbots #99892

Closed
Closed
test_unicodedata: test_normalization() fails randomly with IncompleteRead on PPC64LE Fedora buildbots#99892
Labels
type-bugAn unexpected behavior, bug, or error

Description

@vstinner
vstinner
opened on Nov 30, 2022

For a few weeks, test_unicodedata.test_normalization() fails randomly with IncompleteRead on PPC64LE Fedora buildbots.

IMO the test should be skipped on download error, rather than treating a download error as a test failure.

Example: https://buildbot.python.org/all/#/builders/33/builds/3031

FAIL: test_normalization (test.test_unicodedata.NormalizationTest.test_normalization)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/http/client.py", line 591, in _read_chunked
    value.append(self._safe_read(chunk_left))
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/http/client.py", line 632, in _safe_read
    raise IncompleteRead(data, amt-len(data))
http.client.IncompleteRead: IncompleteRead(8113 bytes read, 2337 more expected)

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/test/test_unicodedata.py", line 362, in test_normalization
    testdata = open_urlresource(TESTDATAURL, encoding="utf-8",
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/test/support/__init__.py", line 671, in open_urlresource
    s = f.read()
        ^^^^^^^^
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/gzip.py", line 295, in read
    return self._buffer.read(size)
           ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/_compression.py", line 118, in readall
    while data := self.read(sys.maxsize):
                  ^^^^^^^^^^^^^^^^^^^^^^
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/gzip.py", line 500, in read
    buf = self._fp.read(READ_BUFFER_SIZE)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/gzip.py", line 90, in read
    return self.file.read(size)
           ^^^^^^^^^^^^^^^^^^^^
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/http/client.py", line 459, in read
    return self._read_chunked(amt)
           ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/http/client.py", line 597, in _read_chunked
    raise IncompleteRead(b''.join(value)) from exc
http.client.IncompleteRead: IncompleteRead(106486 bytes read)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/buildbot/buildarea/3.x.cstratak-fedora-rawhide-ppc64le.clang/build/Lib/test/test_unicodedata.py", line 368, in test_normalization
    self.fail(f"Could not retrieve {TESTDATAURL}")
AssertionError: Could not retrieve http://www.pythontest.net/unicode/15.0.0/NormalizationTest.txt

Linked PRs

Activity

  1. vstinner
    added
    type-bugAn unexpected behavior, bug, or error
    on Nov 30, 2022

  2. vstinner commented on Dec 5, 2022

    @vstinner
    vstinner
    on Dec 5, 2022
    MemberAuthor

    Local patch to trigger the bug:

    diff --git a/Lib/http/client.py b/Lib/http/client.py
index 15c5cf634c..08a0c3b39a 100644
--- a/Lib/http/client.py
+++ b/Lib/http/client.py
@@ -561,6 +561,7 @@ def _get_chunk_left(self):
             if chunk_left is not None:
                 # We are at the end of chunk, discard chunk end
                 self._safe_read(2)  # toss the CRLF at the end of the chunk
+            raise IncompleteRead(b'')
             try:
                 chunk_left = self._read_next_chunk_size()
             except ValueError:
@@ -632,8 +633,7 @@ def _safe_readinto(self, b):
         """Same as _safe_read, but for reading into a buffer."""
         amt = len(b)
         n = self.fp.readinto(b)
-        if n < amt:
-            raise IncompleteRead(bytes(b[:n]), amt-n)
+        raise IncompleteRead(bytes(b[:n]), amt-n)
         return n
 
     def read1(self, n=-1):
@@ -687,8 +687,7 @@ def _read1_chunked(self, n):
             n = chunk_left # if n is negative or larger than chunk_left
         read = self.fp.read1(n)
         self.chunk_left -= len(read)
-        if not read:
-            raise IncompleteRead(b"")
+        raise IncompleteRead(b"")
         return read
 
     def _peek_chunked(self, n):

    Command to reproduce the bug using the patch:

    rm -f ./Lib/test/data/NormalizationTest.txt && ./python -m test -u all test_unicodedata -v
  3. vstinner
    added a commit that references this issue on Dec 5, 2022

    gh-99892: test_unicodedata: skip test on download failure (#100011)

    2488c1e
  4. miss-islington
    added 4 commits that reference this issue on Dec 5, 2022

    pythongh-99892: test_unicodedata: skip test on download failure (pythâ€Ķ

    fccc115

    pythongh-99892: test_unicodedata: skip test on download failure (pythâ€Ķ

    4521373

    gh-99892: test_unicodedata: skip test on download failure (GH-100011)

    c067023

    gh-99892: test_unicodedata: skip test on download failure (GH-100011)

    4a7612f

  5. zware commented on Dec 5, 2022

    @zware
    zware
    on Dec 5, 2022
    Member

    I've started wondering lately if we'd be better served to migrate those test files into the repository.

  6. vstinner commented on Dec 5, 2022

    @vstinner
    vstinner
    on Dec 5, 2022
    MemberAuthor

    I've started wondering lately if we'd be better served to migrate those test files into the repository.

    How many files are we talking about? NormalizationTest.txt takes 2.6 MB.

    My notes on these files: https://pythondev.readthedocs.io/infra.html#services-used-by-unit-tests

    Another example: Lib/test/test_codecmaps_cn.py downloads http://www.pythontest.net/unicode/EUC-CN.TXT which takes 104 KB.

  7. vstinner commented on Dec 5, 2022

    @vstinner
    vstinner
    on Dec 5, 2022
    MemberAuthor

    Files containing pythontest.net/unicode/ pattern:

    • Lib/test/test_codecmaps_cn.py
    • Lib/test/test_codecmaps_hk.py
    • Lib/test/test_codecmaps_jp.py
    • Lib/test/test_codecmaps_kr.py
    • Lib/test/test_codecmaps_tw.py
    • Lib/test/test_ucn.py
    • Lib/test/test_unicodedata.py
    Lib/test/test_codecmaps_cn.py:12:    mapfileurl = 'http://www.pythontest.net/unicode/EUC-CN.TXT'
Lib/test/test_codecmaps_cn.py:17:    mapfileurl = 'http://www.pythontest.net/unicode/CP936.TXT'
Lib/test/test_codecmaps_cn.py:22:    mapfileurl = 'http://www.pythontest.net/unicode/gb-18030-2000.xml'
Lib/test/test_codecmaps_hk.py:12:    mapfileurl = 'http://www.pythontest.net/unicode/BIG5HKSCS-2004.TXT'
Lib/test/test_codecmaps_jp.py:12:    mapfileurl = 'http://www.pythontest.net/unicode/CP932.TXT'
Lib/test/test_codecmaps_jp.py:28:    mapfileurl = 'http://www.pythontest.net/unicode/EUC-JP.TXT'
Lib/test/test_codecmaps_jp.py:35:    mapfileurl = 'http://www.pythontest.net/unicode/SHIFTJIS.TXT'
Lib/test/test_codecmaps_jp.py:49:    mapfileurl = 'http://www.pythontest.net/unicode/EUC-JISX0213.TXT'
Lib/test/test_codecmaps_jp.py:56:    mapfileurl = 'http://www.pythontest.net/unicode/SHIFT_JISX0213.TXT'
Lib/test/test_codecmaps_kr.py:12:    mapfileurl = 'http://www.pythontest.net/unicode/CP949.TXT'
Lib/test/test_codecmaps_kr.py:18:    mapfileurl = 'http://www.pythontest.net/unicode/EUC-KR.TXT'
Lib/test/test_codecmaps_kr.py:28:    mapfileurl = 'http://www.pythontest.net/unicode/JOHAB.TXT'
Lib/test/test_codecmaps_tw.py:12:    mapfileurl = 'http://www.pythontest.net/unicode/BIG5.TXT'
Lib/test/test_codecmaps_tw.py:17:    mapfileurl = 'http://www.pythontest.net/unicode/CP950.TXT'
Lib/test/test_ucn.py:179:        url = ("http://www.pythontest.net/unicode/%s/NamedSequences.txt" %
Lib/test/test_unicodedata.py:359:        TESTDATAURL = f"http://www.pythontest.net/unicode/{unicodedata.unidata_version}/{TESTDATAFILE}"

    I don't know who manages http://www.pythontest.net/unicode/ nor how it is updated to a new Unicode version.

    @benjaminp wrote:

    Basically, I update the version in the script (Tools/unicode/makeunicodedata.py), run it, and fix what fails.

  8. vstinner commented on Dec 6, 2022

    @vstinner
    vstinner
    on Dec 6, 2022
    MemberAuthor

    Issue fixed by: 2488c1e

    Please open a new issue if you want to include these Unicode TXT files directly into Python Git repository.

  9. vstinner
    closed this as completedon Dec 6, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    type-bugAn unexpected behavior, bug, or error

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions