[asterisk-dev] [Code Review] ADSI and Crypto backwards compatibility with 1.6.2

[Tilghman Lesher]

> On 2010-08-25 16:42:39, Kevin Fleming wrote:
> > /branches/1.8/include/asterisk/module.h, lines 245-248
> > <https://reviewboard.asterisk.org/r/876/diff/1/?file=12309#file12309line245>
> >
> >     Make this field already exist; changing the composition of an API structure based on platform dependencies seems very unsafe.
> 
> Tilghman Lesher wrote:
>     Technically, they're only required when the condition applies; otherwise, they need not exist.  I think if we remove the preprocessor definition, the element should be renamed, to avoid confusion... maybe "required_when_gcc_is_broken"?
> 
> Russell Bryant wrote:
>     I think a comment to explain how it is used will suffice.

Made the existence of this field non-conditional in diff r2; also renamed the field.


- Tilghman


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/876/#review2630
-----------------------------------------------------------


On 2010-08-26 15:07:41, Tilghman Lesher wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/876/
> -----------------------------------------------------------
> 
> (Updated 2010-08-26 15:07:41)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Summary
> -------
> 
> In 1.6.2 and previous, res_adsi and res_crypto used stub files to allow them not to be loaded.  We got rid of the stubs in favor of the optional_api approach.  Now people who are upgrading, who previously noload'ed those modules AND are on a platform which does not support optional_api (i.e. gcc 4.1) (e.g. RHEL, Centos 5) are complaining that the easy upgrade path is broken.
> 
> This patch forces those broken platforms to automatically load the dependencies, when required by a loaded module.
> 
> 
> This addresses bug 17707.
>     https://issues.asterisk.org/view.php?id=17707
> 
> 
> Diffs
> -----
> 
>   /branches/1.8/apps/app_adsiprog.c 283524 
>   /branches/1.8/apps/app_followme.c 283524 
>   /branches/1.8/apps/app_getcpeid.c 283524 
>   /branches/1.8/apps/app_queue.c 283524 
>   /branches/1.8/apps/app_speech_utils.c 283524 
>   /branches/1.8/apps/app_stack.c 283524 
>   /branches/1.8/apps/app_voicemail.c 283524 
>   /branches/1.8/channels/chan_agent.c 283524 
>   /branches/1.8/channels/chan_dahdi.c 283524 
>   /branches/1.8/channels/chan_iax2.c 283524 
>   /branches/1.8/channels/chan_mgcp.c 283524 
>   /branches/1.8/channels/chan_sip.c 283524 
>   /branches/1.8/funcs/func_aes.c 283524 
>   /branches/1.8/include/asterisk/module.h 283524 
>   /branches/1.8/main/loader.c 283524 
>   /branches/1.8/pbx/pbx_dundi.c 283524 
>   /branches/1.8/pbx/pbx_loopback.c 283524 
>   /branches/1.8/pbx/pbx_realtime.c 283524 
> 
> Diff: https://reviewboard.asterisk.org/r/876/diff
> 
> 
> Testing
> -------
> 
> noloaded modules on a Centos 5 machine, verified they loaded when a dependent module was loaded, either at boot time, or later, at CLI request (i.e. the 'module load' command).
> 
> 
> Thanks,
> 
> Tilghman
> 
>