Repository navigation
BBC: b3f68b479c breaks combination of Tie::IxHash and YAML::PP #24302
Description
Alerted by this report from @ericherman on the P5P list, I ran this bisection invocation just now on an unthreaded build on Linux:
perl Porting/bisect.pl \
--module=Tie::IxHash,YAML::PP \
--start=v5.43.8
This pointed to b3f68b4 as the first breaking commit:
b3f68b479c9876437620503fb7d54e14bceadcee is the first bad commit
commit b3f68b479c9876437620503fb7d54e14bceadcee
Author: Richard Leach <richardleach@users.noreply.github.com>
Date: Thu Feb 5 00:01:10 2026 +0000
Commit: Richard Leach <rich+perl@hyphen-dash-hyphen.info>
CommitDate: Tue Mar 10 23:46:51 2026 +0000
Convert OP_CONST hash keys in key+val lists to HEKs
...
The breakage lies in YAML-PP's test suite:
Running make test for TINITA/YAML-PP-v0.39.0.tar.gz
PERL_DL_NONLAZY=1 "/home/jkeenan/testing/blead/bin/perl" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(0, 'blib/lib', 'blib/arch')" t/*.t
t/00.compile.t ............. ok
t/10.parse-valid.t ......... 1/? # OK: 318 DIFF: 0 ERROR: 0 TODO: 0 SKIP: 8
t/10.parse-valid.t ......... ok
t/11.parse-invalid.t ....... 1/? # OK: 0 DIFF: 0 ERROR: 96 TODO: 0 SKIP: 5
t/11.parse-invalid.t ....... ok
t/12.load-json.t ........... 1/? # OK: 276 DIFF: 0 ERROR: 0 TODO: 0 SKIP: 8
t/12.load-json.t ........... ok
t/13.load-anchor.t ......... ok
t/14.load-bool.t ........... ok
t/15.parse-eol.t ........... ok
t/16.loader.t .............. ok
t/17.load-complex-keys.t ... ok
t/18.control.t ............. ok
t/19.file.t ................ ok
t/20.dump.t ................ 306/? # OK: 311 DIFF: 0 ERROR: 0 TODO: 0 SKIP: 15
t/20.dump.t ................ ok
t/21.emit.t ................ 1/? # SAME_EVENTS: 283 SAME_YAML: 272 DIFF_EVENTS: 0 DIFF_YAML: 0 ERROR: 0 TODO: 0 SKIP: 43
t/21.emit.t ................ ok
t/22.dump-bool.t ........... ok
t/23-dump-anchor.t ......... ok
t/24.double-escapes.t ...... ok
t/30.legacy.t .............. ok
t/31.schema.t .............. 1/? # inf: Inf -inf: -Inf nan: NaN
t/31.schema.t .............. ok
t/32.cyclic-refs.t ......... ok
t/34.emit-scalar-styles.t .. ok
t/35.highlight.t ........... ok
t/36.debug.t ............... ok
t/37.schema-catchall.t ..... ok
t/37.schema-perl.t ......... # TEST QR: (?^:TEST_STRINGYFY_REGEX)
# QR PREFIX: (?^:
t/37.schema-perl.t ......... ok
t/38.schema-ixhash.t ....... 1/?
# Failed test 'order: dump_string()'
# at t/38.schema-ixhash.t line 30.
# got: '---
# - &1
# U: 2
# B: 52
# c: 64
# '19': 84
# Disco: 2000
# Year: 2525
# days_on_earth: 20000
# - *1
# '
# expected: '---
# - &1
# U: 2
# B: 52
# c: 64
# 19: 84
# Disco: 2000
# Year: 2525
# days_on_earth: 20000
# - *1
# '
# Failed test 'order_blessed: dump_string()'
# at t/38.schema-ixhash.t line 30.
# got: '---
# - &1 !perl/hash:Order
# U: 2
# B: 52
# c: 64
# '19': 84
# Disco: 2000
# Year: 2525
# days_on_earth: 20000
# - *1
# '
# expected: '---
# - &1 !perl/hash:Order
# U: 2
# B: 52
# c: 64
# 19: 84
# Disco: 2000
# Year: 2525
# days_on_earth: 20000
# - *1
# '
# Looks like you failed 2 tests of 2.
t/38.schema-ixhash.t ....... Dubious, test returned 2 (wstat 512, 0x200)
Failed 2/2 subtests
t/39.emitter-alias.t ....... ok
t/40.representers.t ........ ok
t/41.custom.schema.t ....... ok
t/42.tokens.t .............. ok
t/43.indent.t .............. 4/? # SAME_EVENTS: 837 SAME_YAML: 0 DIFF_EVENTS: 0 DIFF_YAML: 0 ERROR: 0 TODO: 0 SKIP: 141
t/43.indent.t .............. ok
t/44.writer.t .............. ok
t/45.binary.t .............. ok
t/46.line-endings.t ........ ok
t/47.header-footer.t ....... ok
t/48.merge.t ............... ok
t/49.include.t ............. ok
t/50.clone.t ............... ok
t/51.directives.t .......... ok
t/52.preserve.t ............ ok
t/53.customtag-alias.t ..... ok
t/54.glob.t ................ ok
t/55.flow.t ................ YAML::PP::Test:216 $skipped = \{
'87E4' => 1,
'6BFJ' => 1,
'CN3R' => 1,
'LQZ7' => 1,
'FRK4' => 1,
'QF4Y' => 1,
'LX3P' => 1,
'4FJ6' => 1,
'4ABK' => 1,
'CT4Q' => 1,
'CFD4' => 1,
'L9U5' => 1,
'8UDB' => 1,
'9MMW' => 1,
'Q9WF' => 1
};
t/55.flow.t ................ 1/? # SAME_EVENTS: 15 SAME_YAML: 0 DIFF_EVENTS: 0 DIFF_YAML: 0 ERROR: 0 TODO: 0 SKIP: 3
t/55.flow.t ................ ok
t/56.force-flow.t .......... 78/? # SAME_EVENTS: 302 SAME_YAML: 0 DIFF_EVENTS: 0 DIFF_YAML: 0 ERROR: 0 TODO: 0 SKIP: 24
t/56.force-flow.t .......... ok
t/57.dup-keys.t ............ ok
Test Summary Report
-------------------
t/38.schema-ixhash.t (Wstat: 512 (exited 2) Tests: 2 Failed: 2)
Failed tests: 1-2
Non-zero exit status: 2
Files=44, Tests=7466, 7 wallclock secs ( 0.26 usr 0.06 sys + 4.71 cusr 0.46 csys = 5.49 CPU)
Result: FAIL
Both Tie::IxHash and YAML::PP are non-core modules, but I have not yet analyzed their interrelationship with respect to these test failures.
@richardleach, can you take a look? Thanks.
Activity
jkeenan commentedon Mar 20, 2026
If we test Tie::IxHash and YAML::PP separately against blead, each PASSes.
I bisected with these two invocations:
In each case, I got
Runner returned 0 for end revision at Porting/bisect.pl line 233.-- where the "end revision" is HEAD (2f3d9f5).richardleach commentedon Mar 20, 2026
The difference seems to be
'19': 84vs19: 84, which does reflect the SV being used as a hash key being hekified.Question is whether this is a functional bug or a more of a change to "dump()" output formatting. If it's the latter, options could include:
S_check_alt_hash_fields_hekifyto only hekify strings, not numbers (even though those numbers will be stringified when used as hash keys).richardleach commentedon Mar 20, 2026
Made a start on exploring this option.
book commentedon Mar 20, 2026
I believe this is because the
YAML-PPtest suite skips the specific test script that is failing isTie-IxHashis not installed.richardleach commentedon Mar 21, 2026
This seems to address the test failure in YAML::PP without that module needing to make any changes.
I'm going to add a separate commit to #24290, since that's already touching the same code.
jkeenan commentedon Mar 21, 2026
I spent a large part of today dealing with the fallout from the perl-5.43.9 release yesterday. I installed that release on both FreeBSD and Linux -- but in this post I will refer only to the CPANtesters results generated on Linux. The routine I use each month installs from the downloaded tarball (unthreaded on Linux), installs Task::CPAN::Reporter against that
perl, then feeds a list of over 500 modules to./bin/cpan. These modules are tested and (if PASSing) installed in dependency ("CPAN River") order. So Tie::IxHash gets tested and installed long before YAML::PP. But once the program records a failure in YAML::PP'st/38.schema-ixhash.t, that module does not get installed -- and all of its dependencies get discarded and no report for them is sent to cpantesters.org. Those modules include these:But these distributions have hundreds of other CPAN distros dependent on them. @richardleach, in the development of your pull request you might wish to test at least the top-level of YAML::PP's reverse dependencies against your branch. You could try Test::Against::Commit to that end.
richardleach commentedon Mar 22, 2026
Thanks for all that effort, @jkeenan. I'll do some testing of the repos you have identified against #24290 later today.
richardleach commentedon Mar 22, 2026
All of these pass using #24290, with the dependency exception of
Net::SSLeay, which had to be force installed. I'll look into that later today.richardleach commentedon Mar 22, 2026
Same
Net::SSLeayfailures with blead, will try 5.43.9 next.jkeenan commentedon Mar 22, 2026
Here is a work-around for Net::SSLeay.
perlagainst which you want to test. For me, recently, this has been aperlbuilt from the perl-5.43.9.tar.gz. For you, it will probably be from your local branch in the p.r. you're testing. Let's call its top-level directory as: "$TESTINGDIR/perl-5.43.9".cd-ed into it.$THISCPANto install the trial version of Net::SSLeay against the monthly dev release.cd $THISDIR; ./bin/cpan Tie::IxHash YAML::PP Dist::ZillaYMMV.
richardleach commentedon Mar 22, 2026
Thanks, I mainly just wanted to check that Net::SSLeay's test failures were unrelated. (They definitely are.)
S_check_alt_hash_fields_hekify - hekified numbers retain valid IV/NVs