If we're in backport season, I have a quick patch that I whipped up
yesterday to fix mod_disk_cache on 2.2.x on systems that have
APR_SENDFILE_ENABLED but EnableSendfile Off.
Issac
William A. Rowe, Jr. wrote:
> [EMAIL PROTECTED] wrote:
>> Votes and Notes; the first four I would like to see applied before
>> tagging 1.2.4; add a better patch for the winnt mpm wait patch
>> that makes this reviewable (sometimes -U3 just isn't enough context.)
>
> As Ruediger guessed and committed (thanks!) we are now two more patches down.
>
> htcacheclean and httxt2dbm was a considerable flaw (considered that they
> didn't behave as documented, and these utilities are supposed to be easily
> moved around from box to box or run from a remote machine) and I'm about
> to commit that.
>
> I see two final state bugs... I'm about to finish vetting niq's patch for
> listen reconfig which pegs httpd at 100% cpu
>
> * core: PR#37680
> Fix NONBLOCK status of listening sockets on restart/graceful
> http://svn.apache.org/viewvc?view=rev&revision=487901
> +1: niq
>
> someone please add a third voice on Nick's patch? I'm guessing nobody
> bothered, next time http://issues.apache.org/bugzilla/show_bug.cgi?id=37680
> would help so the fruit hangs even lower.
>
> The other fatal bug is:
>
>> * mpm_winnt: Fix return values from wait_for_many_objects.
>> http://svn.apache.org/viewvc?view=rev&revision=428029
>> 2.2.x version of patch:
>> Trunk version works
>> + http://people.apache.org/~wrowe/mpm_winnt_waits.patch
>> + is easier to read (-U8)
>> +1: mturk, wrowe
>
> Which i've created another, easier-to-read example of the backport, if you
> didn't follow the change last time please read the new patch and chime in
> a third +1? You'll note it's simply a matter of unsetting errno before we
> invoke the wait, so we know if the 64th, 128th, etc are real errors or in
> fact indexes into the wait list array.
>
> That's it. I'd like to roll tonight so the vote can be closed by the 22nd.
> If another patch today fixes a runaway cpu or platform fault I might fix,
> if it's behavioral/optimization I won't be holding for it. Those can go in
> the next 2 - 6 weeks whenever Jim rolls the release he's been aspiring to.
Index: modules/cache/mod_disk_cache.c
===================================================================
--- modules/cache/mod_disk_cache.c (revision 488297)
+++ modules/cache/mod_disk_cache.c (working copy)
@@ -359,6 +359,10 @@
cache_info *info;
disk_cache_object_t *dobj;
int flags;
+#if APR_HAS_SENDFILE
+ core_dir_config *pdconf = ap_get_module_config(r->per_dir_config,
+ &core_module);
+#endif
h->cache_obj = NULL;
@@ -453,7 +457,8 @@
/* Open the data file */
flags = APR_READ|APR_BINARY;
#ifdef APR_SENDFILE_ENABLED
- flags |= APR_SENDFILE_ENABLED;
+ flags |= ((pdconf->enable_sendfile == ENABLE_SENDFILE_OFF)
+ ? 0 : APR_SENDFILE_ENABLED);
#endif
rc = apr_file_open(&dobj->fd, dobj->datafile, flags, 0, r->pool);
if (rc != APR_SUCCESS) {
