git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 3/3] read-cache: don't write index twice if we can't write shared index

From
SZEDER Gábor <szeder.dev@gmail.com>
Date
Jan 22, 2018, 18:27 UTC
Message-ID
<20180122182717.21539-1-szeder.dev@gmail.com>
In-Reply-To
<CACsJy8BBXQ9KErfiuf2ty_4szE2fiHLDiKvMig1LbSefzf-o7w@mail.gmail.com>
On Thu, Jan 18, 2018 at 1:47 PM, Duy Nguyen <pclouds@gmail.com> wrote:
Show 23 quoted lines
> On Thu, Jan 18, 2018 at 6:36 PM, SZEDER Gábor <szeder.dev@gmail.com> wrote:
>> This series, queued as 'nd/shared-index-fix', makes the 32 bit Linux
>> build job fail on Travis CI.  Unfortunately, all it can tell us about
>> the failure is this:
>>
>>   Test Summary Report
>>   -------------------
>>   t1700-split-index.sh                             (Wstat: 256 Tests: 23
>>   Failed: 1)
>>     Failed test:  23
>>     Non-zero exit status: 1
>>   Files=809, Tests=18491, 401 wallclock secs ( 7.22 usr  1.60 sys + 263.16
>>   cusr 49.58 csys = 321.56 CPU)
>>   Result: FAIL
>>
>> because it can't access the test's verbose log due to lack of
>> permissions.
>>
>>
>>   https://travis-ci.org/git/git/jobs/329681826#L2074
>
> I may need help getting that log (or even better the trash directory
> of t1700). 

Well, shower thoughts gave me an idea, see the included PoC patch below. I can't really decide whether it's too clever or too ugly :)

It produces output like this (a previous WIP version without '--immediate'):

  https://travis-ci.org/szeder/git/jobs/331601009#L2486
  -- >8 --
Subject: [PATCH] travis-ci: include the trash directories of failed tests in
 the trace log

The trash directory of a failed test might contain valuable information about the cause of the failure, but we have no access to the trash directories of Travis CI build jobs. The only feedback we get from there is the trace log, so...

Modify 'ci/print-test-failures.sh' to create a tar.gz archive of the test directory of each failed test and encode it with base64, so the result is a block of ASCII text that can be safely included in the trace log, along with a hint about how to restore it. Furthermore, run tests with '--immediate' to faithfully preserve the failed state.

A few of our tests create a sizeable trash directory, so limit the size of each included base64-encoded block, let's say, to 1MB.

Note:
  - The logs of Linux build jobs coming from Travis CI have mostly
    CRLF line endings, which makes 'base64 -d' from 'coreutils'
    complain about "invalid input"; it has to be converted to LF
    first.  'openssl base64 -d' can grok it just fine, even without
    conversion.
  - The logs of OSX build jobs have CRCRLF line endings.  However, the
    'base64' util of OSX doesn't wrap its output at 76 columns, i.e.
    prints one single albeit very long line.  This means that while
    'base64' from 'coreutils' still complains, by the time it gets to
    the invalid input it already decoded everything and produced a
    valid .tar.gz.  OTOH, 'openssl base64 -d' doesn't grok it, but
    exits without any error message or even an error code, even after
    converting to CRLF or LF line endings.
    Go figure.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
---
 ci/lib-travisci.sh        |  2 +-
 ci/print-test-failures.sh | 20 ++++++++++++++++++++
 2 files changed, 21 insertions(+), 1 deletion(-)
diff --git a/ci/lib-travisci.sh b/ci/lib-travisci.sh
index 1efee55ef7..981c6e9504 100755
--- a/ci/lib-travisci.sh
+++ b/ci/lib-travisci.sh
@@ -97,7 +97,7 @@ fi
 export DEVELOPER=1
 export DEFAULT_TEST_TARGET=prove
 export GIT_PROVE_OPTS="--timer --jobs 3 --state=failed,slow,save"
-export GIT_TEST_OPTS="--verbose-log"
+export GIT_TEST_OPTS="--verbose-log --immediate"
 export GIT_TEST_CLONE_2GB=YesPlease
 
 case "$jobname" in
diff --git a/ci/print-test-failures.sh b/ci/print-test-failures.sh
index 4f261ddc01..cc6a1247a4 100755
--- a/ci/print-test-failures.sh
+++ b/ci/print-test-failures.sh
@@ -23,5 +23,25 @@ do
 		echo "$(tput setaf 1)${TEST_OUT}...$(tput sgr0)"
 		echo "------------------------------------------------------------------------"
 		cat "${TEST_OUT}"
+
+		TEST_NAME="${TEST_EXIT%.exit}"
+		TEST_NAME="${TEST_NAME##*/}"
+		TRASH_DIR="t/trash directory.$TEST_NAME"
+		TRASH_TGZ_B64="$TEST_NAME.trash.base64"
+		if [ -d "$TRASH_DIR" ]
+		then
+			tar czp "$TRASH_DIR" |base64 >"$TRASH_TGZ_B64"
+
+			if [ 1048576 -lt $(wc -c <"$TRASH_TGZ_B64") ]
+			then
+				# larger than 1MB
+				echo "$(tput setaf 1)Trash directory of '$TEST_NAME' is too big to be included in the trace log$(tput sgr0)"
+			else
+				echo "$(tput setaf 1)Trash directory of '$TEST_NAME':$(tput sgr0)"
+				echo "(Extract it to a file from the raw log, make sure it has Unix line endings, then run 'base64 -d <file> |tar vxzp' to restore it)"
+				cat "$TRASH_TGZ_B64"
+				echo "$(tput setaf 1)End of trash directory of '$TEST_NAME'$(tput sgr0)"
+			fi
+		fi
 	fi
 done
-- 
2.16.1.80.gc0eec9753d
Previous: Jeff KingNext: Eric Sunshine
Message 87 of 95 in “fixes for split index mode”
  1. 0/3 fixes for split index modeThomas Gummerer, Dec 10, 2017
  2. 1/3 repository: fix repo_read_index with submodulesThomas Gummerer, Dec 10, 2017
  3. Brandon WilliamsDec 11, 2017
  4. Thomas GummererDec 11, 2017
  5. 2/3 prune: fix pruning with multiple worktrees and split indexThomas Gummerer, Dec 10, 2017
  6. Brandon WilliamsDec 11, 2017
  7. Thomas GummererDec 11, 2017
  8. 3/3 travis: run tests with GIT_TEST_SPLIT_INDEXThomas Gummerer, Dec 10, 2017
  9. Eric SunshineDec 10, 2017
  10. SZEDER GáborDec 11, 2017
  11. Thomas GummererDec 11, 2017
  12. Lars SchneiderDec 12, 2017
  13. Junio C HamanoDec 12, 2017
  14. Thomas GummererDec 12, 2017
  15. Junio C HamanoDec 12, 2017
  16. Thomas GummererDec 13, 2017
  17. Lars SchneiderDec 13, 2017
  18. Junio C HamanoDec 13, 2017
  19. Lars SchneiderDec 13, 2017
  20. Thomas GummererDec 13, 2017
  21. 0/3 fixes for split index modeThomas Gummerer, Dec 17, 2017
  22. 2/3 prune: fix pruning with multiple worktrees and split indexThomas Gummerer, Dec 17, 2017
  23. Brandon WilliamsDec 18, 2017
  24. Thomas GummererJan 3, 2018
  25. Junio C HamanoJan 4, 2018
  26. 1/3 repository: fix repo_read_index with submodulesThomas Gummerer, Dec 17, 2017
  27. Brandon WilliamsDec 18, 2017
  28. Thomas GummererDec 18, 2017
  29. Brandon WilliamsDec 18, 2017
  30. 3/3 travis: run tests with GIT_TEST_SPLIT_INDEXThomas Gummerer, Dec 17, 2017
  31. Lars SchneiderDec 18, 2017
  32. Thomas GummererJan 4, 2018
  33. Lars SchneiderJan 5, 2018
  34. Thomas GummererJan 7, 2018
  35. 0/3 fixes for split index modeThomas Gummerer, Jan 7, 2018
  36. 1/3 read-cache: fix reading the shared index for other reposThomas Gummerer, Jan 7, 2018
  37. Duy NguyenJan 8, 2018
  38. Thomas GummererJan 8, 2018
  39. Thomas GummererJan 13, 2018
  40. Brandon WilliamsJan 8, 2018
  41. Duy NguyenJan 9, 2018
  42. Brandon WilliamsJan 16, 2018
  43. Duy NguyenJan 17, 2018
  44. Brandon WilliamsJan 17, 2018
  45. Jonathan NiederJan 17, 2018
  46. Duy NguyenJan 18, 2018
  47. Junio C HamanoJan 19, 2018
  48. Thomas GummererJan 20, 2018
  49. Junio C HamanoJan 22, 2018
  50. Thomas GummererJan 27, 2018
  51. Junio C HamanoFeb 7, 2018
  52. 2/3 split-index: don't write cache tree with null oid entriesThomas Gummerer, Jan 7, 2018
  53. 3/3 travis: run tests with GIT_TEST_SPLIT_INDEXThomas Gummerer, Jan 7, 2018
  54. 4/3 read-cache: don't try to write index if we can't write shared indexThomas Gummerer, Jan 13, 2018
  55. Duy NguyenJan 14, 2018
  56. 1/3 read-cache.c: change type of "temp" in write_shared_index()Nguyễn Thái Ngọc Duy, Jan 14, 2018
  57. 2/3 read-cache.c: move tempfile creation/cleanup out of write_shared_indexNguyễn Thái Ngọc Duy, Jan 14, 2018
  58. 3/3 read-cache: don't write index twice if we can't write shared indexNguyễn Thái Ngọc Duy, Jan 14, 2018
  59. SZEDER GáborJan 18, 2018
  60. Duy NguyenJan 18, 2018
  61. Jeff KingJan 18, 2018
  62. Duy NguyenJan 18, 2018
  63. Duy NguyenJan 18, 2018
  64. Jeff KingJan 18, 2018
  65. SZEDER GáborJan 18, 2018
  66. Duy NguyenJan 19, 2018
  67. 0/5 Travis CI: don't run the test suite as root in the 32 bit Linux buildSZEDER Gábor, Jan 22, 2018
  68. 2/5 travis-ci: use 'set -e' in the 32 bit Linux build jobSZEDER Gábor, Jan 22, 2018
  69. Jeff KingJan 23, 2018
  70. Jeff KingJan 23, 2018
  71. SZEDER GáborJan 24, 2018
  72. Jeff KingJan 24, 2018
  73. 5/5 travis-ci: don't fail if user already exists on 32 bit Linux build jobSZEDER Gábor, Jan 22, 2018
  74. Jeff KingJan 23, 2018
  75. Duy NguyenJan 24, 2018
  76. SZEDER GáborJan 24, 2018
  77. 1/5 travis-ci: use 'set -x' for the commands under 'su' in the 32 bit Linux buildSZEDER Gábor, Jan 22, 2018
  78. 3/5 travis-ci: don't repeat the path of the cache directorySZEDER Gábor, Jan 22, 2018
  79. Jeff KingJan 23, 2018
  80. SZEDER GáborJan 24, 2018
  81. 4/5 travis-ci: don't run the test suite as root in the 32 bit Linux buildSZEDER Gábor, Jan 22, 2018
  82. Jeff KingJan 23, 2018
  83. SZEDER GáborJan 24, 2018
  84. Jeff KingJan 24, 2018
  85. Jeff KingJan 24, 2018
  86. Jeff KingJan 24, 2018
  87. SZEDER GáborJan 22, 2018
  88. Eric SunshineJan 22, 2018
  89. SZEDER GáborJan 22, 2018
  90. Duy NguyenJan 24, 2018
  91. Lars SchneiderJan 26, 2018
  92. Thomas GummererJan 14, 2018
  93. Thomas GummererJan 18, 2018
  94. Junio C HamanoJan 19, 2018
  95. Thomas GummererJan 19, 2018

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.