- FreeSWITCH-dev - FreeSWITCH Mailing Lists

[PATCH] fixing conference dial flexibly, fixing bad pointer op
by Travis Cross 22 Oct '08

22 Oct '08

Greetings, I noticed that the following currently fails in FreeSWITCH: freeswitch@domain> conference 3001 dial {foo=bar}sofia/internal/2001%domain 3001 XML default Conference 3001 30 2008-10-22 04:01:14 [ERR] switch_core_session.c:249 switch_core_session_outgoing_channel() Could not locate channel type {foo=bar}sofia 2008-10-22 04:01:14 [ERR] switch_ivr_originate.c:1025 switch_ivr_originate() Cannot create outgoing channel of type [{foo=bar}sofia] cause: [CHAN_NOT_IMPLEMENTED] API CALL [conference(3001 dial {foo=bar}sofia/internal/2001%domain 3001 XML default Conference 3001 30)] output: Call Requested: result: [CHAN_NOT_IMPLEMENTED] The example looks contrived, but this functionality is fairly critical when you need to add, for example, P-Asserted-Identity or Diversion headers for the call to go outbound. Interestingly, I found that conference bgdial works correctly, which got me looking at the source code. Here is the offender, conference_outcall(...) in mod_conference.c: char *dialstr = switch_mprintf("{ignore_early_media=true}%s", bridgeto); status = switch_ivr_originate(NULL, &peer_session, cause, dialstr, 60, NULL, cid_name, cid_num, NULL, SOF_NONE); This means that {ignore_early_media=true}{foo=bar}sofia/... is getting passed to switch_ivr_originate. Initially I looked for a better way to pass channel variables to switch_ivr_originate, but I didn't see a good way to do it without changing the interface around. Then I thought, why shouldn't switch_ivr_originate simply accept multiple sets of braces? This feature could be useful in other contexts: scripts, for example, that need to add channel variables and then include a dialstring from another application layer, just like conference_outcall is doing. So I came up with the following patch, which also fixes a bug in the current code that I'll note below the patch: --- Author: Travis Cross <tc(a)traviscross.com> Date: Sat Oct 18 00:40:05 2008 +0000 Allow multiple sets of channel variable brackets on originate. This change is intended in part to prevent channel variables that we add in the C code from interfering with the user's ability to use this feature. This occurs, for example, in conference_outcall. diff --git a/src/switch_ivr_originate.c b/src/switch_ivr_originate.c index 02c1a37..57e002b 100644 --- a/src/switch_ivr_originate.c +++ b/src/switch_ivr_originate.c @@ -577,6 +577,8 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_originate(switch_core_session_t *sess switch_call_cause_t reason = SWITCH_CAUSE_NONE; uint8_t to = 0; char *var_val, *vars = NULL; + int var_block_count = 0; + char *e = NULL; const char *ringback_data = NULL; switch_codec_t *read_codec = NULL; uint8_t sent_ring = 0, early_ok = 1, return_ring_ready = 0, progress = 0; @@ -611,18 +613,41 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_originate(switch_core_session_t *sess data++; } - if (*data == '{') { - char *e = switch_find_end_paren(data, '{', '}'); - - if (e) { + /* extract channel variables, allowing multiple sets of braces */ + while (*data == '{') { + if (!var_block_count) { + e = switch_find_end_paren(data, '{', '}'); + if (!e || !*e) { + goto var_extract_error; + } vars = data + 1; - *e++ = '\0'; - data = e; + *e = '\0'; + data = e + 1; } else { - switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Parse Error!\n"); - status = SWITCH_STATUS_GENERR; - goto done; + if (e) { + *e = ','; + } + e = switch_find_end_paren(data, '{', '}'); + if (!e || !*e) { + goto var_extract_error; + } + /* swallow the opening bracket */ + int i = 0, j = 0; + while ((data + i) && *(data + i)) { + j = i; i++; + /* note that this affects vars[] */ + data[j] = data[i]; + } + *(--e) = '\0'; + data = e + 1; } + var_block_count++; + continue; + + var_extract_error: + switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Parse Error!\n"); + status = SWITCH_STATUS_GENERR; + goto done; } /* strip leading spaces (again) */ --- After applying the patch, conference dial works as expected. It is now supported and well defined to use multiple contiguous sets of curly brackets in a dial string to pass channel variables. The patch also corrects the following code from the original source: char *e = switch_find_end_paren(data, '{', '}'); if (e) { vars = data + 1; *e++ = '\0'; data = e; If data does not contain a closing bracket, then switch_find_end_paren will return the address of the end of the string. The code above would then set data to the first address in memory after the end of the string, which would be very undefined. Cheers, -- tc

2 2

Win32 Missing apr_iconv.h
by David Brazier 22 Oct '08

22 Oct '08

Compiling on Win32 from SVN 10107, VS complained that it couldn't find apr_iconv.h, referenced from libs/apr-util/xlate/xlate.c: #if APU_HAVE_APR_ICONV #include <apr_iconv.h> #endif So I changed apu.h: #define APU_HAVE_APR_ICONV 0 // was 1 And now it compiles OK. But maybe that wasn't the right thing to do? I notice apr_iconv.h was in SVN (libs/apr-iconv-1.1.1/include) and was deleted back in May. David 360CRM Limited, registered address, Caledonia House, 89 Seaward Street, Glasgow, G41 1HJ, Registered in Scotland, company number SC261952. This email and its attachments may be confidential and are intended solely for the use of the individual to whom it is addressed. Any views or opinions expressed are solely those of the author and do not necessarily represent those of 360CRM Ltd. This email does not form any part of a contract. If you are not the intended recipient of this email and its attachments, you must take no action based upon them, nor must you copy or show them to anyone. Please contact the sender if you believe you have received this email in error.

2 1

Notes about Windows build of svn 10088
by jlenk＠frontiernet.net 21 Oct '08

21 Oct '08

If you have an existing checkout/build you must manually modify apu.h from libs\apr-util\include with the modifications from apu.hw in svn10085 Attached patch for trivial warning in mod_xml_rpc Also if you use mod_managed you must reswig. Thanks, Jeff

2 2

[PATCH] Fixed loop unrolling in dialplan configuration
by Travis Cross 17 Oct '08

17 Oct '08

Greetings, While grepping through the sources to identify how channel variables are used (and I must say, the FreeSWITCH sources are exquisitely beautiful C code), I uncovered the following minor issue. It appears this would have prevented loop unrolling from working as intended with the in-tree dialplan. Author: Travis Cross <tc(a)traviscross.com> Date: Fri Oct 17 03:54:12 2008 +0000 Fixed loop unrolling in dialplan configuration. unroll_loops is a preprocessor variable, not a channel variable. diff --git a/conf/dialplan/default.xml b/conf/dialplan/default.xml index 61a7500..6bf4ca3 100644 --- a/conf/dialplan/default.xml +++ b/conf/dialplan/default.xml @@ -15,7 +15,7 @@ <context name="default"> <extension name="unloop"> - <condition field="${unroll_loops}" expression="^true$"/> + <condition field="$${unroll_loops}" expression="^true$"/> <condition field="${sip_looped_call}" expression="^true$"> <action application="deflect" data="${destination_number}"/> </condition> diff --git a/conf/dialplan/public.xml b/conf/dialplan/public.xml index 49b8850..5c8cdc5 100644 --- a/conf/dialplan/public.xml +++ b/conf/dialplan/public.xml @@ -14,7 +14,7 @@ <context name="public"> <extension name="unloop"> - <condition field="${unroll_loops}" expression="^true$"/> + <condition field="$${unroll_loops}" expression="^true$"/> <condition field="${sip_looped_call}" expression="^true$"> <action application="deflect" data="${destination_number}"/> </condition> Cheers, -- Travis

2 3

ECHO
by apparao gsp 13 Oct '08

13 Oct '08

Hi I am newbie, I was installed successfylly freeswitch , and also run the mod_perl files but how to interface between mod_perl and freeswitch and I have an error of to the default commands of *echo* like this 2008-10-13 15:07:00 [CONSOLE] switch_console.c:244 switch_console_process() Unknown Command: echo so plz clarufy my doubt with regards Apparao GSP

2 1

XCF-2820
by Mircea Carasel 13 Oct '08

13 Oct '08

Hi, Regarding: http://track.sipfoundry.org/browse/XCF-2820 This ticket requests to change the conference original path where the config files are written (/usr/local/freeswitch/conf/). Config files (conference.conf.xml and default_context.xml) should be relocated in the sipXconfig default config path (/etc/sipxpbx) Following a discussion with Damian, we think that freeswitch needs some changes in order to look for these files in the new config path Please share your comments Thanks & Regards, Mircea Carasel

2 1

Critical module load (abort FS on failure)
by Michael Giagnocavo 12 Oct '08

12 Oct '08

Hello, I wrote a tiny feature to allow configuration of a module as critical. This is done in the modules.conf.xml file by adding critical="true" to any load element. It works by making FreeSWITCH completely fail module loading (and hence exit) if any critical module does not return true. The idea is that if FS fails to load something important, then the FS process will exit. This allows you to know very immediately when something failed versus figuring it out at runtime. Any comments in general? Also, should this be set on any modules in the default configs? In particular, I was thinking some of the other language plugins that are commented out. If someone uncomments them, it's because they want to use it. Hence, if it fails, it should probably fail right in their face by exiting. Should other modules have this set by default? (Sofia?) Thanks, -Michael

2 1

Quick OZ Question: physical disconnect/reconnect of PRI
by Michael Collins 11 Oct '08

11 Oct '08

Guys, I've been using FS to troubleshoot a carrier issue. I need to roll patch cables as I try different circuits. I noticed that when I physically disconnect a PRI and plug it back in that the dchan stays down. A shutdown/restart of FS brings it back up. Q1: Is there a way to have OZ try to bring the dchan back up after a layer 1 disconnect/reconnect? Q2: Would it be horribly difficult to add a command like "oz bounce <span number>" for the sake of testing a circuit? (That is, doing a soft restart on the span instead of bringing FS all the way down and back up.) Thanks! -MC

3 2

Shadow copy for mod_managed
by Michael Giagnocavo 10 Oct '08

10 Oct '08

Hello, I'm thinking of adding a ShadowCopy function to mod_managed to allow easy overwriting and reloading of managed modules. Here's the comment description for the feature. Any comments? Shadow copy loading The build in ShadowCopy loading for an AppDomain is nice, but it will not reload an assembly with the same path. So when overwriting assemblies, the Load/LoadFrom doesn't help and just returns the currently loaded assembly. (LoadFile bypasses Fusion, and hence the ShadowCopy feature). Because of this, we implement our own shadow copy loading facility. On initial load, all files in the managed module dir are copied to the shadow dir under the "0" folder and loaded from there. (Existing files in the shadow folder are deleted.) When loading a new assembly after the initial load, it is copied to the shadow dir, under the current version (start at 1, go up). Resolved dependencies are also copied over to the current version folder. Thanks, Michael

2 2

SS7
by Evgeniy Zolotov 10 Oct '08

10 Oct '08

Greetings for all ! By us has been developed, as FreeSWITCH module, module for connecting under protocol SS7 based on library libss7 Asterisk Open Source PBX. By first we have made this as separate module mod_ss7, but futher we plan to make it as a part of mod_openzap, like it has been done for ISDN. Our module provides levels MTP2 and MTP3 of SS7. Now we work over support of ISUP. If this line of development is of interest for You - we ready to give results and initial codes. Faithfully, Evgeniy.

3 2