summary refs log tree commit diff
path: root/contrib
diff options
context:
space:
mode:
authorJune McEnroe <june@causal.agency>2020-11-16 15:04:39 -0500
committerJune McEnroe <june@causal.agency>2020-11-16 15:04:39 -0500
commit09af6dcd618701bc5994b4a146e6df34e1cf9015 (patch)
treecbc26cac234aaf730cdd169736b02f8ee58c75e3 /contrib
parentSwap localAccept parameter order (diff)
downloadpounce-09af6dcd618701bc5994b4a146e6df34e1cf9015.tar.gz
pounce-09af6dcd618701bc5994b4a146e6df34e1cf9015.zip
Set client sockets non-blocking
Except for during writes. This prevents pounce getting blocked on a
client sending only a partial TLS record, for example.

Writes still need to block because pounce doesn't have a way to resume
them. (And it would do so by having a buffer, but sockets already have a
send buffer, so what would be the point of that?) I don't think it
should be a problem since outside of stateSync, writes only happen when
poll returns POLLOUT. I feel like ideally SO_SNDLOWAT would be set to
guarantee a full IRC message can always be written on POLLOUT, but since
it's actually TLS records being sent, it's not obvious what the size
would be.

I'm also making an assumption here that tls_read returning
TLS_WANT_POLLOUT is unlikely to happen, since I don't actually set
pollfd.events based on that. I'm not sure how wanting to resume a
tls_read after a POLLOUT could be cleanly handled. I'm just going to
hope that if it does happen, the regular poll loop will eventually sort
it out...
Diffstat (limited to '')
0 files changed, 0 insertions, 0 deletions
class='logmsg'> This actually speeds things up quite a bit, saving roughly a second on a big PNG screenshot. Almost all the remaining time is spent in deflate. 2021-09-21Rewrite pngo, add explicit optionsJune McEnroe Interesting to see how my code habits have changed. 2021-09-16Fix /* **/ comment matchingJune McEnroe 2021-09-15Remove typer, add downgrade to READMEJune McEnroe 2021-09-15Set bot mode on downgradeJune McEnroe 2021-09-15Enter capsicum in downgradeJune McEnroe 2021-09-15Factor out common parts of downgrade messagesJune McEnroe Also bump the message cap to 1024 because that is ostensibly useful for replying to older messages. 2021-09-14Add downgrade IRC botJune McEnroe 2021-09-14Sort by title if authors matchJune McEnroe There are probably better things to sort by but title definitely always exists. 2021-09-13Swap-remove tags as they're foundJune McEnroe This makes it even faster. From ~1s on a sqlite3.c amalgamation to ~0.85s. 2021-09-12Replace htagml regex with strncmpJune McEnroe Since ctags only ever produces regular expressions of the form /^re$/ or /^re/ with no other special characters, instead unescape the pattern and simply use strncmp. Running on a sqlite3.c amalgamation, the regex version takes ~37s while the strncmp version takes ~1s, producing identical output. Big win! 2021-09-11Also defer printing comment for lone close-parensJune McEnroe 2021-09-10Publish "git-comment"June McEnroe 2021-09-10Add git comment --pretty optionJune McEnroe 2021-09-08Defer printing comment if line is blank or closing braceJune McEnroe This fixes badly indented comments. 2021-09-08Up default min-repeat to 30 linesJune McEnroe 2021-09-08Handle dirty lines in git-commentJune McEnroe 2021-09-08Document and install git-commentJune McEnroe 2021-09-08Add repeat and all options to git-commentJune McEnroe 2021-09-08Add group threshold to git-commentJune McEnroe