Bug #88724 for podlators: Insufficient sanity check

Shouldn’t this be in the podlaters queue? That’s the distribution that includes Pod::Man. https://rt.cpan.org/Public/Dist/Display.html?Name=podlators Best, David

Show quoted text
Maybe, but run the snippet with Carp::Always and see who was the last caller: perl -e ' use Pod::Man; use Carp::Always; for my $i (128){ # pick any number 128..255 my $content = "=".chr($i); eval { Pod::Man->new->parse_string_document($content) }; printf "%d: %s\n", $i, $@; }' 128: Modification of non-creatable array value attempted, subscript -1 at /home/sand/src/perl/repoperls/installed-perls/perl/v5.19.3/a2da/lib/5.19.3/Pod/Man.pm line 287. Pod::Man::_handle_element_start('Pod::Man=HASH(0x1af58a0)', 'head1', 'HASH(0x1b0a1f0)') called at /home/sand/src/perl/repoperls/installed-perls/perl/v5.19.3/a2da/lib/site_perl/5.19.3/Pod/Simple/BlackBox.pm line 1484 Pod::Simple::BlackBox::_traverse_treelet_bit(undef, undef, 'HASH(0x1b0a1f0)', 'POD ERRORS') called at /home/sand/src/perl/repoperls/installed-perls/perl/v5.19.3/a2da/lib/site_perl/5.19.3/Pod/Simple/BlackBox.pm line 880 Pod::Simple::BlackBox::_ponder_paragraph_buffer('Pod::Man=HASH(0x1af58a0)') called at /home/sand/src/perl/repoperls/installed-perls/perl/v5.19.3/a2da/lib/site_perl/5.19.3/Pod/Simple/BlackBox.pm line 76 Pod::Simple::BlackBox::parse_lines('Pod::Man=HASH(0x1af58a0)', undef) called at /home/sand/src/perl/repoperls/installed-perls/perl/v5.19.3/a2da/lib/site_perl/5.19.3/Pod/Simple.pm line 405 Pod::Simple::parse_string_document('Pod::Man=HASH(0x1af58a0)', '=\x{80}') called at -e line 6 eval {...} called at -e line 6 Downgrading the Pod::Simple 3.20 makes this bug disappear.

On Wed Sep 18 00:23:13 2013, david@justatheory.com wrote: Show quoted text
David asked me to look at the issue on GitHub since I was responsible for the commit referenced above. If I understand correctly, the problem being reported here is that this 1-liner: perl -MPod::Man -e 'Pod::Man->new->parse_string_document("=\x80")' will die with this error: Modification of non-creatable array value attempted, subscript -1 at lib/Pod/Man.pm ... The condition that is triggering this failure is indeed the new parse warning introduced in my commit, however the same result can be triggered on versions of Pod::Simple that pre-date my change, with this 1-liner: perl -MPod::Man -e 'Pod::Man->new->parse_string_document("=cut")' The problem seems to relate to the fact that both these snippets of POD source trigger parse warnings but are otherwise empty. Under these circumstances, Pod::Simple will generate an errata section and perhaps Pod::Man is not in a suitable state to handle the errata section if it hasn't seen any preceding document elements (wild guess on my part). Regards Grant

Could podlators please comment on the nature of this bug and what needs to be fixed? Thanks!

On 2013-09-18 12:25:19, DWHEELER wrote: Show quoted text
Ah, already moved over to the podlater queue. Sorry. David

On 2013-09-18 03:32:12, ANDK wrote: Show quoted text
This will be fixed in the next release of podlators. The problem was that, for these documents, Pod::Simple sets the flag indicating that they're contentless. In an ill-advised meaningless optimization, Pod::Man (and Pod::Text) didn't set up internal state for output of documents that are contentless. However, in this case, the supposedly contentless document still has a POD ERRORS section and content, so subsequent processing referenced data structures that weren't initialized. I think it's a little weird that Pod::Simple claims a document with content (even manufactured content) is contentless, but there may be some reason to keep that behavior so that callers can still treat empty documents with errors as empty documents. Regardless, it's now fixed in my development tree by removing the ill-advised and useless optimization and doing full setup for empty documents.

Fixed in 2.5.2, now released. Note that the test I added made the mistake of assuming that this error would always occur, so it fails on all Perls with older Pod::Simple. I will release 2.5.3 to fix that.