This Week on perl5-porters (16-22 June 2003)

This Week on perl5-porters (16-22 June 2003)

This week, there was fewer threads than usual, but they were larger, so the volume of mail exchanged on p5p remained more or less constant. Here's the high signal-to-noise ratio summary for this week.

Algorimic Complexity Attack

You might remember that last month, Scott Crosby reported (as bug #22371) a way to produce colliding hash keys. Once fed into a hash, these may severely degrade performance of element lookup, and thus be used as a basis for DoS attacks. Jarkko proposed a way to protect hash tables against this : instead of using 0 as initial value for the computation of the hash value, as it is now, use some pseudo-random value, computed at interpreter startup time.

This raises two interesting questions : is this a significant threat ? And how to enhance the hash function without performance degradation ? Hopefully the arguments ended in some interesting benchmarks. To be continued.

Clearing $1

Dan Kogai thinks that it would be a good idea, if a failing match did reset the $<digit> variables to undef. This opinion is not universally shared. SeveralLots of arguments are proposed against this change, from backwards compatibility to good coding practices.


Jarkko Hietaniemi proposed an experimental patch to take advantage of the POSIX signal handling feature that goes by the name of siginfo. It enables additional information to be passed to signal handlers : sending process id, real user id of sending process, etc. (See your sigaction(2) manpage if you have it.)

Selected bugs

Perl segfaults during make install on Mario A Cruz Gartner's FreeBSD 4.4 box (bug #22740). That's puzzling because make and make test both succeeded.

Shlomi Fish and Ton Hospel reported bug #22744, demonstrating an old problem with the .= operator :

    $b = ($a.="5")."6"; print "$a $b"

prints "56 56", but it should have been "5 56". Enache Adrian provided a fix.

About this summary

This week's short summary was written by Rafael Garcia-Suarez. Weekly summaries are available on and/or via a mailing list, which subscription address is Comments, corrections, additions, and suggestions are welcome.