httpd

Checkout Tools
  • last updated 47 mins ago
Constraints
Constraints: committers
 
Constraints: files
Constraints: dates

Changeset 1824136 is being indexed.

Installation instructions, and links, for the majority of people who

will not be building from source.

Fixes weird formatting issue in the output.

hostgroup patch v1.0
    • ?
    /httpd/patches/trunk/hostgroup-v1.0.diff
mod_proxy_fcgi: prioritize the check for mayflush when using flushpackets

The mayflush variable should be checked before the rest

to avoid polling when not needed.

Suggested by Yann Ylavic on the dev@ mailing list.

More corrections to mod_ssl Spanish translation
documentation rebuild
mod_proxy_fcgi.xml: add more context to the conn reuse note
documentation rebuild
mod_proxy_fcgi.xml: add more context to the conn reuse note
Axe some code duplication.

Use 'ap_array_str_contains()' instead of 'ctutil_in_array()'

Rebuild.

XML updates.

Rebuild.

XML updates.

On the 2.4.x branch:

Merged /httpd/httpd/trunk:r1821371,1822502-1822503,1822624

  1. … 40 more files in changeset.
promote
Propose
Rebuilt
Add missing idtype="section"
rebuild
Fix a broken link + fix some layout to synch with trunk
documentation rebuild
mod_proxy_fcgi.xml: add a note about connection reuse
documentation rebuild
mod_proxy_fcgi.xml: add a note about connection reuse

Fix a broken link and redirect to the online IRC client
backport prop vote

Merge r1821558, r1821561, r1821595 from trunk:

mod_event: Let the listener thread do its maintenance job on resources shortage.

PR 61979.

Follow up to r1821558: CHANGES typo.

mpm_event: Follow up to r1821558.

Don't crash (in listener) if we can't create the ptrans allocator.

Submitted by: ylavic

Reviewed by: ylavic, jim, icing

Merge r1819855, r1821562 from trunk:

mpm_event: wakeup the listener to re-enable listening sockets.

When listening sockets are disabled (too many connections) and the number of

workers / active connections comes back below the limit, we need to wake up

the listener to re-enable them.

Add a new connections_above_limit() helper to determine when this applies.

Follow up to r1819855: CHANGES entry.

Submitted by: ylavic

Reviewed by: ylavic, jim, icing

Merge r1643279, r1703241, r1802535, r1819847, r1819848, r1819852, r1819853 from trunk:

mpm_event(opt): avoid casts/comparisons from unsigned to signed (atomics).

mpm_event/worker: make ap_queue_term() atomic (acquire/release the mutex once).

mpm_event: ap_queue_info_try_get_idler() may atomically decrement and then

re-increment the number idlers if it went under or to zero. We can avoid

this by switching to a compare-and-swap scheme.

mpm_event: avoid unexpected compiler optimizations.

Make sure the compiler doesn't play games with our synchronization variables

by marking them volatile.

mpm_event: make sure wakeup_listener() does its minimal job.

Even if the listener thread is not created yet (i.e. about to be), we must

still tell it to leave, and terminate the worker queue in any case.

mpm_event: worker factor vs pollset.

Make sure the worker factor is at least one (w.r.t. WORKER_FACTOR_SCALE), and

use it to size the pollset appropriately (including K-A and lingering close

connections), in addition to the listening sockets.

mpm_event: remove atomics for timeout_queue's total counter.

It's always updated under the timeout_mutex lock, or read for logging and

scoreboard updates (not critical).

For the read cases a volatile access is enough, while removing the atomic ops

for the already protected write cases saves cycles and context switches.

Submitted by: ylavic

Reviewed by: ylavic, jim, icing