RFR: 8357510: [REDO] RunTest variables should always be assigned
Magnus Ihse Bursie
ihse at openjdk.org
Wed May 28 00:02:04 UTC 2025
This is a redo of [JDK-8357048](https://bugs.openjdk.org/browse/JDK-8357048), which had to backed out since it caused testing errors in higher tiers.
The problem was that `JTREG_PROBLEM_LIST_PREFIX` was not defined before it was used, and when `JTREG_BASIC_OPTIONS` were no longer implicitly declared as a macro, but instead got a definite assignment, the value of JTREG_PROBLEM_LIST_PREFIX was empty at the time of evaluation.
I have now manually checked each and every `+=` assignment to `$1_JTREG_BASIC_OPTIONS`, and verified that all variables present is defined earlier.
Here is the original description from JBS:
When building `$1_JTREG_BASIC_OPTIONS`, it is assumed that the variable is recursively defined and that thus `+=` is lazy.
$1_JTREG_BASIC_OPTIONS += -$$($1_JTREG_TEST_MODE) \
-verbose:$$(JTREG_VERBOSE) -retain:$$(JTREG_RETAIN) \
-concurrency:$$($1_JTREG_JOBS) -timeoutFactor:$$(JTREG_TIMEOUT_FACTOR) \
-vmoption:-XX:MaxRAMPercentage=$$($1_JTREG_MAX_RAM_PERCENTAGE) \
-vmoption:-Dtest.boot.jdk="$$(BOOT_JDK)" \
-vmoption:-Djava.io.tmpdir="$$($1_TEST_TMP_DIR)"
```
If `+=` is eagerly evaluated, the option -timeoutFactor: will get an empty argument and fail.
The problem is the line: `$$(eval $$(call SetJtregValue,$1,JTREG_BASIC_OPTIONS))` might create the variable `$1_JTREG_BASIC_OPTIONS` "simply expanded" (the expansion will create an assignment using `:=`). Whereas if the variable is not created the first `+=` will be recursive and will work as expected.
One solution to this problem is replacing the three assignments in `SetJtregValue` from `:=` to `=`. This might have other side effects.
A more conservative solution might be to create another macro (thus not changing behaviour where strict evaluation might be needed):
define SetJtregRecursiveValue
ifneq ($$($2), )
$1_$2 = $$($2)
else
ifneq ($$($$($1_COMPONENT)_$2), )
$1_$2 = $$($$($1_COMPONENT)_$2)
else
ifneq ($3, )
$1_$2 = $3
endif
endif
endif
endef
-------------
Commit messages:
- Fix bug with original implementation
- Revert "8357511: [BACKOUT] 8357048: RunTest variables should always be assigned"
Changes: https://git.openjdk.org/jdk/pull/25475/files
Webrev: https://webrevs.openjdk.org/?repo=jdk&pr=25475&range=00
Issue: https://bugs.openjdk.org/browse/JDK-8357510
Stats: 18 lines in 1 file changed: 11 ins; 6 del; 1 mod
Patch: https://git.openjdk.org/jdk/pull/25475.diff
Fetch: git fetch https://git.openjdk.org/jdk.git pull/25475/head:pull/25475
PR: https://git.openjdk.org/jdk/pull/25475
More information about the build-dev
mailing list