BBC: POWERMAN/Crypt-MatrixSSL3-v3.9.3.tar.gz: leak test may fail

[eserte]

t/leak.t may fail since 5.41.something like this:

#   Failed test 'MEM: newkeys_deletekeys'
#   at t/leak.t line 270.
#     '336'
#         <=
#     '288'
# Looks like you failed 1 test of 26.
t/leak.t .................................. 
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/26 subtests 

However, there are also mixed results here. For 5.41.4, both pass and fail reports exist:

• http://www.cpantesters.org/cpan/report/6842daba-8206-11ef-b2ad-38246748ae50 (pass)
• http://www.cpantesters.org/cpan/report/d4bd6670-8e35-11ef-b6e9-5e516e8775ea (fail)

Note also that the module has a different issue with -DDEBUGGING perls, so https://fast2-matrix.cpantesters.org/?dist=Crypt-MatrixSSL3%20v3.9.3 shows also fail reports with older perl versions.

[jkeenan]

t/leak.t may fail since 5.41.something like this:

#   Failed test 'MEM: newkeys_deletekeys'
#   at t/leak.t line 270.
#     '336'
#         <=
#     '288'
# Looks like you failed 1 test of 26.
t/leak.t .................................. 
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/26 subtests 

However, there are also mixed results here. For 5.41.4, both pass and fail reports exist:

I'm getting similarly inconsistent results with my CPANtesters reports (http://fast-matrix.cpantesters.org/?dist=Crypt-MatrixSSL3;os=linux;reports=1#sl=6,1). Bisection has been frustrated by runs that report the end version passing. Not sure how to proceed.

[jkeenan]

Note also that the module has a different issue with -DDEBUGGING perls, so https://fast2-matrix.cpantesters.org/?dist=Crypt-MatrixSSL3%20v3.9.3 shows also fail reports with older perl versions.

The issue on -DDEBUGGING builds was reported by @andk upstream at powerman/perl-Crypt-MatrixSSL3#42 and was bisected to 6ba07ce from back in 2015:

commit 6ba07ce12b615587e55caf4477539992d4b3e79b
Author:     Jarkko Hietaniemi <[email protected]>
AuthorDate: Fri Aug 7 22:14:55 2015 -0400
Commit:     Jarkko Hietaniemi <[email protected]>
CommitDate: Fri Aug 14 15:49:58 2015 -0400

    No __attribute__((nonnull(...))) from NN.

(I have trouble building perls at commits that old, so I haven't been able to confirm the bisection.)

Andreas pointed to a CPANtesters failure on a threaded debugging build run in 2022. If we were to resolve that older failure it might give us a clue to dealing with the current one.

[richardleach]

t/leak.t may fail since 5.41.something like this:

#   Failed test 'MEM: newkeys_deletekeys'
#   at t/leak.t line 270.
#     '336'
#         <=
#     '288'
# Looks like you failed 1 test of 26.
t/leak.t .................................. 
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/26 subtests 

This might also be partially 06e421c - test code no longer leaking as before. I'll take a look after putting in a PR for #22704

[richardleach]

This might also be partially 06e421c

It is. More simulated leaking of a string buffer. PR submitted as powerman/perl-Crypt-MatrixSSL3#45

[jkeenan]

This might also be partially 06e421c

It is. More simulated leaking of a string buffer. PR submitted as powerman/perl-Crypt-MatrixSSL3#45

This pull request was accepted by the upstream maintainer and Crypt-MatrixSSL3-v3.9.4 was released on Nov 5. I tested that on my CPANtesters rigs where the tests were previously failing. They are now passing, so this ticket can be closed. Doing so now.