httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rob Hartill <hart...@ooo.lanl.gov>
Subject 0.7.2 uploaded
Date Tue, 06 Jun 1995 20:43:23 GMT

Okay kiddies, as promised, there's a new version of the pre-forking
Apache on hyperreal, called 0.7.2

changes..

  a few more globals move into the per-request data structure
  including the document_root - hopefully (untested by me) it'll be
  reset after each request. Other VIRTUAL HOST globals will need the
  same treatment later.

  I added my nph replacement prototype.. scripts now have CLIENT_FD
  to use or ignore. If it sucks (can't see why it should) it can be
  removed.

  Fixed cgi external redirect defaut content-type to be text/html so
  as to match up with the HTML Apache adds by default.

  Added "connection with client broke" error log message - until I
  find out how to pass the per-request data structure to a signal
  handler (without using yucky global variables), the client address
  is missing :-(

  The parent process no longer goes to sleep after giving birth to
  the children, now it sits waiting for any of them to die (*) so
  it can spawn a replacement.

  (*) Child processes now self destruct after processing 
  conf/httpd.conf   "MaxRequestsPerChild" requests (built-in default
  is 40 I think)
  This should minimise damage  due to leaks.

  Because Apache children service requests in a round-robin fashion,
  they can all hit MaxRequestsPerChild at the same time on a quiet
  server. A busy server will (proabably) tie up enough of the children
  for long enough to miss their next turn, and thus cause the self 
  destructs to be randomly distributed. Explicit randomness can be 
  introduced if need be.

Go break it...

Now what ?,

I can carry on moving more globals into the per-request data structure,
or we can all join in and start uploading patches.


--
Rob Hartill                           
http://nqcd.lanl.gov/~hartill/

Mime
View raw message