Fix three minor issues with gzipped manual page files.
1. In the <title> element, disregard a .gz suffix when constructiog
the name(section) string; implemented by rewriting the file name
parsing code in resp_begin_html().
2. If a query returns exactly one result, strip any .gz suffix
from the HTTP redirect destination.
3. If a search returns more than one result, strip any .gz suffixes
from the A HREF destinations.
These issues were found by Abigail G <dev at placeviolette.net>,
but my patch differs from hers in so far as it does not need
additional memory allocations and only touches two functions,
pg_searchres() and resp_begin_html().
throttle reads from the peer socket if more than MSG_PROCESS_LIMIT messages
are ready.
Only MSG_PROCESS_LIMIT messages are process once during a poll loop so it
makes no sense to buffer more messages by keeping the POLLIN enabled.
Once enough messages have been processed the POLLIN is enabled again
and more messages are read in. This should prevent excessive buffering
in the session engine.
OK tb@
rpki-client: mop up a few more struct parse remnants
This drops uses of p for signed objects outside of printing. Mechanical
diff apart from the addition of a line break to avoid an overlong line
in roa_read().
ok claudio
Fix edge case where aspa_table_equal() may return equal even though the
tables are not the same.
Just comparing customer-as and the flat array of all provider-as sets
does not detect if the right AS switches between two neighboring
customer-as sets. Since the customer-as did not change and the flat
array remains the same.
Additionally check the number of provider-as per customer-as to detect
this edge condition.
Reported by N0zoM1z0
OK tb@
