This Week on perl5-porters (3-9 February 2003)

This Week on perl5-porters (3-9 February 2003)

I'm not sure whether it's me or if the Perl 5 porters discussions were actually a bit shorter than usual, but this week's summary, indeed, appears to be a bit shorter than usual. This doesn't mean that it's absolutely without interest.

Maintenance releases

Leon Brocard asked for the expected availability of the 5.005_04 and 5.6.2 releases. Notably, there are known fixes to allow building perl 5.005_03 and 5.6.1 with the newest versions of gcc and of the Berkeley DB. Gurusamy Sarathy intends to release a 5.6.2, and adds that smoke tests results are welcome. Tim Bunce and Nicholas Clark add that it's important that the Perl community be seen to actively support at least one previous major version.

Reducing op size

Stephen McCamant provided some ideas about removing the op_seq field from the op node structure (more precisely, to reduce it to two flags). An op could be reasonably identified by a couple (address, checksum of fields). He also sent a patch for B::Concise to remove its dependency on op_seq.

Seeking a safe signal test

Jarkko Hietaniemi proposes to add a new magic variable ${^SIGNAL_UNSAFE} (or something) to restore locally the pre-5.8.0 "unsafe" signal behavior when set to a true value.

He asks whether someone could provide sample code of where the safe signals "reliably" gets it wrong, to be used as a basis for tests for this new feature. To be noted, this kind of tests are extremely dependent on the platform.

In brief

Robert Spier posted a handy list of untouched bugs, alongside his weekly perlbug summary. He says that a bunch of em look easy, and I feel guilty : he's right.

Dustin Jones reports an interesting parser bug (#20716) : the lexical $x in the double-quoted string below is not seen by perl :

    my $x = 'x'; print "${x}{";

(removing the my solves the problem.) Enache Adrian provided a fix.

Dan Kogai reported that formats don't seem to play well with PerlIO and Unicode strings. This didn't surprise Jarkko. Inaba Hiroto proposed some patches to fix this.

Enache Adrian was busy solving several bugs related to goto : segfaults (#20154, #20357) and a newly discovered memory leak (#20777).

Ton Hospel reported three new regular expressions bugs (if I count correctly).

Abigail requests that Test::Harness should provide a way to be optionally silenced. Currently it prints a string for each test ran (the counter.) This may take some time for large test suites.

H. Merijn Brand reports that Cygwin v1.3.20 now has nanosleep() and that bleadperl successfully compiles and passes all tests on it.

About this summary

This summary brought to you by Rafael Garcia-Suarez. Summaries are available on and/or via a mailing list, which subscription address is Comments and corrections are welcome.