Fix parsing of IPv6 addresses #53
Conversation
|
See zeek/zeekscript#106 for the related change there. |
While we are not able to parse dual IPv6 addresses (zeek/zeekscript#34) we apparently were not even able to parse Zeek IPv6 address literals since we did not include the surrounding `[..]` in the grammar. This patch takes care of that, but leaves zeek/zeekscript#34 unaddressed.
Our previous regex for IPv6 addresses was very complicated but still did not accommodate all possible formats. This patch radically simplifies it so it should be able to match all possible addresses at the cost of performing less validation during parsing which seems like a reasonable trade-off. Closes #34.
bbannier
force-pushed
the
topic/bbannier/ipv6
branch
from
5c48f00 to
152a814
Compare
There was a problem hiding this comment.
Thanks for looking into this! This was a total construction site when I left off, per the comment about wanting to revisit Zeek's own regex for v6.
The one you put in accepts things like [:::1] or [abcdef], but I think you're okay with that, right? I agree that that's obviously better than rejecting valid Zeek v6 strings, as long as it doesn't cause conflicts with other [...] constructs.
Yes, this accepts even invalid ipv6 addresses, mainly so we do not turn "parsing" into "full-fledged validation". I deliberately left ipv4 more strict so it does not conflict with parsing of |
Closes #34.