![]() |
SA Bugzilla – Full Text Bug Listing |
Summary: | failure on idn_dots.t | ||
---|---|---|---|
Product: | Spamassassin | Reporter: | Kevin A. McGrail <kmcgrail> |
Component: | Regression Tests | Assignee: | SpamAssassin Developer Mailing List <dev> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | apache, billcole, giovanni, kmcgrail, nicolas.rochelemagne |
Priority: | P2 | ||
Version: | 3.4 SVN branch | ||
Target Milestone: | Undefined | ||
Hardware: | PC | ||
OS: | Windows NT | ||
Whiteboard: | |||
Attachments: |
Patch to reversethe fast untaint in Util.pm
Test if get_uri_list and spamassassin script agree on number of URIs |
Description
Kevin A. McGrail
2018-06-26 15:00:27 UTC
Created attachment 5578 [details]
Test if get_uri_list and spamassassin script agree on number of URIs
Counts the URIs in 6 simple messages 3 ways: direct grep, get_uri_list(), and spamassassin -D
Some more info that could explain the bug in RH bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1597565 Thanks Giovanni. Glad you were able to confirm that a custom compiled perl resolves the issue too. With the removal of the untaint code that was causing the failure, I'm using Bill's tests to see if I think we can release things. But it is good to open a bug with RH as it points to some issue with their compilation is possible. Test passes on CentOS 7.5 with perl 5.16 built from src.rpm (with all RH patches) but with -D_FORTIFY_SOURCE=0. This is definitely a Perl bug that has been fixed in later Perl versions. IMHO the best we can do is revert the offending line of code. Thanks for cross referencing to RH. Any word if RH is able to fix it? I think this clears the blocker with the code removed. RH asked me to recreate the problem with a snippet, without using all SA codebase. I tried a bit but I failed. Reversing the fast untaint patch and considering closed. It's a bug in perl but getting a snippet for RH so they can backport isn't something the project can do. Committed revision 1838387. Committed revision 1838388. we could also consider checking the perl version and if $] is greater than for example 5.025 then we could use this fast untaint technique? The problem is accepting "performance" patches with zero proof of actually delivering. The patch that included this "faster untaint" had several other changes that actually worsened performance. |