Thread (8 messages) 8 messages, 4 authors, 2021-07-19

Re: [PATCH v2] perf: fix when running with TEST_OUTPUT_DIRECTORY

From: Jeff King <hidden>
Date: 2021-07-19 09:25:33

On Fri, Jun 18, 2021 at 03:56:08PM +0200, Patrick Steinhardt wrote:
When the TEST_OUTPUT_DIRECTORY is defined, then all test data will be
written in that directory instead of the default directory located in
"t/". While this works as expected for our normal tests, performance
tests fail to locate and aggregate performance data because they don't
know to handle TEST_OUTPUT_DIRECTORY correctly and always look at the
default location.

Fix the issue by adding a `--results-dir` parameter to "aggregate.perl"
which identifies the directory where results are and by making the "run"
script awake of the TEST_OUTPUT_DIRECTORY variable.
OK, that makes sense. My first thought is that the aggregation script
could simply use $TEST_OUTPUT_DIRECTORY itself, but:

  - we don't actually export that (and changing that is likely to have
    undesirable secondary effects)

  - people may run aggregate separately anyway (I know I do in order to
    produce nice tables without having to pointlessly re-run all tests)

So the new option makes sense. It is a little less convenient when
running aggregate manually (you have to say "--results-dir" manually),
but it's better than not being able to do it at all. :)

A few notes / questions below:
quoted hunk ↗ jump to hunk
diff --git a/t/perf/perf-lib.sh b/t/perf/perf-lib.sh
index 601d9f67dd..f5ed092ee5 100644
--- a/t/perf/perf-lib.sh
+++ b/t/perf/perf-lib.sh
@@ -45,7 +45,7 @@ export TEST_DIRECTORY TRASH_DIRECTORY GIT_BUILD_DIR GIT_TEST_CMP
 MODERN_GIT=$GIT_BUILD_DIR/bin-wrappers/git
 export MODERN_GIT
 
-perf_results_dir=$TEST_OUTPUT_DIRECTORY/test-results
+perf_results_dir=$TEST_RESULTS_DIR
This line puzzled me a bit. Isn't $TEST_RESULTS_DIR already defined to
be $TEST_OUTPUT_DIRECTORY/test-results? If the change here is just for
clarity / readability that's OK by me. I just want to make sure I'm not
missing something.
quoted hunk ↗ jump to hunk
@@ -253,7 +253,10 @@ test_size () {
 # and does it after running everything)
 test_at_end_hook_ () {
 	if test -z "$GIT_PERF_AGGREGATING_LATER"; then
-		( cd "$TEST_DIRECTORY"/perf && ./aggregate.perl $(basename "$0") )
+		(
+			cd "$TEST_DIRECTORY"/perf &&
+			./aggregate.perl --results-dir="$TEST_RESULTS_DIR" $(basename "$0")
+		)
 	fi
 }
OK, and we pass it unconditionally, which should work because it will
always be set, even if you aren't overriding the default. Good.
quoted hunk ↗ jump to hunk
diff --git a/t/perf/run b/t/perf/run
index c7b86104e1..d19dec258a 100755
--- a/t/perf/run
+++ b/t/perf/run
@@ -188,10 +188,10 @@ run_subsection () {
 
 	if test -z "$GIT_PERF_SEND_TO_CODESPEED"
 	then
-		./aggregate.perl $codespeed_opt "$@"
+		./aggregate.perl --results-dir="$TEST_RESULTS_DIR" $codespeed_opt "$@"
 	else
-		json_res_file="test-results/$GIT_PERF_SUBSECTION/aggregate.json"
-		./aggregate.perl --codespeed "$@" | tee "$json_res_file"
+		json_res_file=""$TEST_RESULTS_DIR"/$GIT_PERF_SUBSECTION/aggregate.json"
+		./aggregate.perl --results-dir="$TEST_RESULTS_DIR" --codespeed "$@" | tee "$json_res_file"
 		send_data_url="$GIT_PERF_SEND_TO_CODESPEED/result/add/json/"
 		curl -v --request POST --data-urlencode "json=$(cat "$json_res_file")" "$send_data_url"
 	fi
In the earlier hunks from perf-lib.sh, we got $TEST_RESULTS_DIR by
sourcing test-lib.sh. But in the "run" script, we don't. So who sets
it? Looks like...
quoted hunk ↗ jump to hunk
@@ -203,10 +203,17 @@ get_var_from_env_or_config "GIT_PERF_SEND_TO_CODESPEED" "perf" "sendToCodespeed"
 cd "$(dirname $0)"
 . ../../GIT-BUILD-OPTIONS
 
-mkdir -p test-results
-get_subsections "perf" >test-results/run_subsections.names
+if test -n "$TEST_OUTPUT_DIRECTORY"
+then
+	TEST_RESULTS_DIR="$TEST_OUTPUT_DIRECTORY/test-results"
+else
+	TEST_RESULTS_DIR=test-results
+fi
...we now do. Makes sense. And we will get TEST_OUTPUT_DIRECTORY
directly from GIT-BUILD-OPTIONS (or it won't be set at all if
unspecified).

So the whole thing looks good to me, assuming there is nothing confusing
about the one assignment I mentioned.

-Peff
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help