Revert tilghman and pari's code changes, as
we do NOT need to uri_decode in manager.
(if I sent core%20show%20channels from a telnet
session, it should be interpreted literally, however,
if I send that from an http session, it should be
decoded, which is the behaivor now)
Fix some errant device states by making the devicestate API more strict in
terms of the device argument (only without the unique identifier appended).
(closes issue #12771)
Reported by: davidw
Patches:
20080717__bug12771.diff.txt uploaded by Corydon76 (license 14)
Tested by: davidw, jvandal, murf
Committing a fix that was introduced a long time
ago (does not affect 1.4), where you would pass
a pointer to the end of a character array, and
ast_uri_decode would do no good.
Russell Bryant [Fri, 25 Jul 2008 14:57:11 +0000 (14:57 +0000)]
Modify the main page of the doxygen documentation to link to a new page dedicated
to Asterisk licensing information. The licensing page includes the Asterisk license,
as well as a (not yet complete) list of 3rd party libraries that may be used, as well
as what license we receive them under.
Help filling out this list in the format that I have started in doxyref.h would be
much appreciated. :)
Russell Bryant [Fri, 25 Jul 2008 12:42:25 +0000 (12:42 +0000)]
When the ast_device_state() function is called to retrieve device state, and
the code checks to see if there is a cached state available, use the aggregate
cached state across all servers, and not just the local state.
The 'data' field in the ast_exten struct was being
'moved' from the current dialplan to the replacement
dialplan. This was not good, as the current dialplan
could have problems in the time between the change
and when the new dialplan is swapped in.
So, I modified the merge_and_delete code to strdup
the 'data' field (the args to the app call), and
then it's freed as normal.
I improved a few messages; I added code to limit
the number of calls to the context_merge_incls_swits_igps_other_registrars()
to one per context. I don't think having it called
multiple times per context was doing anything bad,
but it was inefficient.
I hope this fixes the problems Mr. Geiss was noting in
asterisk-users, see
http://lists.digium.com/pipermail/asterisk-users/2008-July/215634.html
As suggested by seanbright, the PSEUDO_CHAN_LEN in
app_chanspy should be set at load time, not at compile
time, since dahdi_chan_name is determined at load time.
Also changed the next_unique_id_to_use to have the
static qualifier.
Also added the dahdi_chan_name_len variable so that
strlen(dahdi_chan_name) isn't necessary. Thanks to
seanbright for the suggestion.
Small cleanup. Move the declaration of the DAHDI_SPANINFO
variable to the block where it is used. This allows one
less #ifdef HAVE_PRI to clutter things up.
Thanks to Tzafrir for pointing this out on #asterisk-dev
minor optimization for stringfields: when a field is being set to a larger value than it currently contains and it happens to be the most recent field allocated from the currentl pool, it is possible to 'grow' it without having to waste the space it is currently using (or potentially even allocate a new pool)
another Fix because of r119585, this commit has broken high frequented BRI Ports, there was a possibility that a channel, that was marked as in_use would be reused later, the corresponding port could got stuck then. So it is recommended to upgrade for chan_misdn users.
........
Mark Michelson [Tue, 22 Jul 2008 22:17:09 +0000 (22:17 +0000)]
Merged revisions 132777 via svnmerge from
https://origsvn.digium.com/svn/asterisk/trunk
........
Allow Spiraled INVITEs to work correctly within Asterisk.
Prior to this change, a spiraled INVITE would cause a 482
Loop Detected to be sent to the caller. With this change,
if a potential loop is detected, the Request-URI is inspected
to see if it has changed from what was originally received. If
pedantic mode is on, then this inspection is fully RFC 3261
compliant. If pedantic mode is not on, then a string comparison
is used to test the equality of the two R-URIs.
This has been tested by using OpenSER to rewrite the R-URI
and send the INVITE back to Asterisk.
fix up namespace pollution for dahdi_chan_mode enum
correct registration of AMI actions in chan_dahdi; in zap-only mode, only register the Zap flavors of the actions (and use Zap prefixes for headers and acks), but in dahdi+zap mode, register both Zap and DAHDI flavors of actions
ensure that if any alarms exist at channel creation time, they are handled identically to if they occurred later, so that later alarm clearing will work properly and 'make sense'
The most common question on the #asterisk iRC channel and on mailing lists
seems to be in regards to an error message when retransmit fails. This
is frequently misunderstood as a failure of Asterisk, not a failure of
the network to reach the other party.
This document tries to assist the Asterisk user in sorting out these
issues by explaining the logic and pointing at some possible
causes. Hopefully, we will get other questions now :-)
Optionally build integer-based routines for FSK tone decoding (but default
to the more accurate float-based routines).
(Closes issue #11679)
(Step 1 of 2)
Fix an issue in iax2 where a call that's been rejected still kept an open channel on the side that attempted to make the call (not the side of the
call that rejected the call). Changes were load tested and also approved by Russell.
Russell Bryant [Mon, 21 Jul 2008 14:47:41 +0000 (14:47 +0000)]
Remove libresample from the Asterisk source tree. It is now available in its
own repository, and must be installed like any other library for Asterisk to
use. The two modules that require it are codec_resample and app_jack.
To install libresample:
$ svn co http://svn.digium.com/svn/libresample/trunk libresample
$ cd libresample
$ ./configure
$ make
$ sudo make install
This code is currently in our own repository because the build system did not
include the appropriate targets for building a dynamic library or for installing
the library.
grant a license exception to allow distribution of Asterisk binaries that use the UW IMAP Toolkit (which is licensed under a non-GPL-compatible license)
Fixes problem where manager users loaded from users.conf would be
removed early (before the routine to load the configuration was
finished) because a variable wasn't initialized.
Russell pointed out that using ast_strdupa() within a loop like this is
probably not a good idea, as we might run out of stack space. Therefore,
changing this over to use the ast_str infrastructure for buffers is
probably a good idea.
Revert part of issue #5620 (revision 6965) as it appears that it was in error.
This should fix talk call progress on analog lines.
(closes issue #12178)
Reported by: michael-fig
Patches:
20080717__bug12178.diff.txt uploaded by Corydon76 (license 14)
Russell Bryant [Thu, 17 Jul 2008 14:46:29 +0000 (14:46 +0000)]
Instead of attempting to pass through AST_EVENT_DEVICE_STATE, use DEVICE_STATE_CHANGE
instead. DEVICE_STATE is a state change on one server, and DEVICE_STATE_CHANGE is
the "real" state of that device across all servers sharing state. This would have
only been a problem with distributed device state.
Steve Murphy [Wed, 16 Jul 2008 23:53:02 +0000 (23:53 +0000)]
(closes issue #13089)
Reported by: murf
Most of this bug was already fixed by Tilghman before
I opened it; Many thanks to Tilghman for his fix
in svn version 125794. That fix cleared up some of the
fields in the lock_info.
This commit changes the address that is stored for the
lock in the lock_info struct, so that it is the same
as that passed into the locking macros. This makes
searching for a lock_info (as in log_show_lock())
by its lock addr possible. The lock_addr field is
infinitely more useful if it is the same as what
is 'publicly' available outside the lock_info code.
Many thanks to kpfleming, putnopvut, and Russell for their
invaluable insights earlier today.
Always ensure that the channel's tech_pvt reference is NULL after calling the
destroy callback.
(closes issue #13060)
Reported by: jpgrayson
Patches:
chan_iax2_tech_pvt_crash.patch uploaded by jpgrayson (license 492)
Move the init_queue call back to where it used to be (changed
Sept 12 last year). It was moved then to prevent a memory leak.
Since then, the same memory leak recurred and was fixed in a
better way.
Now it has been found that the placement of this init_queue
call can cause problems if a realtime queue has values changed
to an empty string. The problem is that the default value
for that queue parameter would not be set.
Make absolutely certain that the transfer datastore
is removed from the calling channel once the caller
is finished in the queue. This could have weird con-
sequences when dialing local queue members when multiple
transfers occur on a single call.
Also fixed a memory leak that would occur when an
attended transfer occurred from a queue member.
The problem was that, esoteric as it is, because the hangerupper
context immediately preceded the std-priv-extent macro, that
the checking code accidentally would fall from traversing hangerupper
into the std-priv-exten macro, where it would hit the hangerupper
in the 'includes', and proceed into an infinite recursion.
A small fix to traverse into the statements of the context instead
of the context solves this issue.
I also added some commented out printfs for debug, which were pretty
handy in the face of a dorky gdb.
This was a problem around since the package was first written;
but evidently pretty rare in turning up in the field.
Steve Murphy [Tue, 15 Jul 2008 23:36:19 +0000 (23:36 +0000)]
(closes issue #12960)
Reported by: mnicholson
Spent most of the day on this bug, and the
solution was so simple. Just had to find and
understand the problem.
The problem was, that the routine to copy
the existing switches, includes, and ignorepats
from the old context to the new one, wasn't
getting called when the context is already
existent. (In other words, if AEL is adding
a new context to the mix, they get copied,
but if pbx_config already defined a context,
then the copy wasn't happening. This made
no sense, so I moved the call to copy the
includes & etc, no matter the case.
Russell Bryant [Tue, 15 Jul 2008 18:46:40 +0000 (18:46 +0000)]
Fix a couple of places in res_agi where the agi_commands lock would not be
released, causing a deadlock. (Reported by mvanbaak in #asterisk-dev,
discovered by bbryant's change to the lock tracking code to yell at you
if a thread exits with a lock still held)
astman_send_error does not need a newline appended -- the API takes care of
that for us.
(closes issue #13068)
Reported by: gknispel_proformatique
Patches:
asterisk_1_4_astman_send.patch uploaded by gknispel (license 261)
asterisk_trunk_astman_send.patch uploaded by gknispel (license 261)
projects / thirdparty / asterisk.git / log
