Fix base.mk recipe loops that ignore a failed iteration

[jnasbyupgrade]

Five base.mk targets run one command per extension or per file, and a failed command was ignored unless it was the last one, so make exited 0 after a real failure. They now fail as soon as any one command fails.

Target	Command run per item	Failure that was ignored
pgtle	pgtle.sh --extension <ext>	any extension but the last
results	cp test/results/<name>.out test/expected/<name>.out	any file but the last
results-build	cp test/build/results/<name>.out test/build/expected/<name>.out	any file, including the last: the skipped-file check after the loop discarded the loop's status
print-pgtle	cat pg_tle/<version>/<ext>.sql	any extension but the last
test/install/schedule (with PGXNTOOL_ENABLE_TEST_INSTALL=yes)	echo "test: ../install/<name>" >> test/install/schedule	any file but the last, leaving an incomplete schedule

pgtle joined its $(foreach ...) calls with ;. It now runs them in a shell for loop over the extensions, with || exit 1 on each call. make -n pgtle in a two-extension project:

for ext in ext_alpha ext_beta; do \
	pgxntool/pgtle.sh --extension $ext  || exit 1; \
done

PGXNTOOL_PGTLE_VERSION handling is unchanged (--pgtle-version 1.5.0+ is still appended to each call when set).

The other four targets put || exit 1 on the command inside the loop.

Tests

Postgres-Extensions/pgxntool-test#108 adds regression tests for pgtle, results and results-build. Each forces a failure by putting a directory where a file would be written, which fails even when the tests run as root. For results, the failing copy is not the last one.

pgtle was also checked by hand in a scratch two-extension project, replacing one extension's pg_tle/1.5.0+/<ext>.sql output with a directory:

Case	Before	After
first extension fails	exit 0	exit 2
last extension fails	exit 2	exit 2
no failure	exit 0	exit 0
no failure, PGXNTOOL_PGTLE_VERSION=1.5.0+	-	exit 0

print-pgtle and the install schedule have no regression test, because neither failure can be forced portably:

• print-pgtle already rejects a missing or non-regular file before cat runs. A test would need cat to fail on a regular file, which means a read error. Checked by hand on Linux, in a scratch two-extension project, by symlinking the first extension's pg_tle/1.5.0+/<ext>.sql to /proc/self/mem. That path is a regular file whose read fails with Input/output error. make -o pgtle print-pgtle PGXNTOOL_PGTLE_TARGET_VERSION=1.5.2 exited 0 before this change and exits 2 after it. /proc is Linux-only, so this isn't a committed test.
• Every append in the schedule loop writes to the file that the line before the loop has just created with >. Anything a test can do to that path also makes that first write fail, and make already stops there. Only a write error partway through the loop, such as a full disk, reaches the ignored case.

Fixes #140

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

• Configuration used: Organization UI
• Review profile: ASSERTIVE
• Plan: Advanced
• Run ID: 032af818-56aa-46c2-b729-3cfa4745d623

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}