[asterisk-dev] [Code Review]: chan_sip code cleanup

[Mark Michelson]

> On June 21, 2012, 3:48 p.m., jrose wrote:
> > /trunk/channels/chan_sip.c, lines 16508-16514
> > <https://reviewboard.asterisk.org/r/1993/diff/1/?file=28966#file28966line16508>
> >
> >     I don't really see why this kind of minor optimization is being grouped with some of the other stuff in this patch. I'm also not to fond of variables with less than useful names like 'ptr' for a character pointer... though if it's going to be recycled like this, there probably isn't a real problem with that. Maybe something like delimiter_ptr or something would be more appropriate than ptr.

I agree, saving 4-8 bytes of stack space is not "cleaning up" the code. Giving a variable like 'ptr' yet another purpose in the function just makes things harder to follow. Not immensely harder, mind you, but still it makes it harder to pin down the purpose of specific variables.


> On June 21, 2012, 3:48 p.m., jrose wrote:
> > /trunk/channels/chan_sip.c, lines 7453-7459
> > <https://reviewboard.asterisk.org/r/1993/diff/1/?file=28966#file28966line7453>
> >
> >     This seems to be a minor optimization (to avoid redeclaring the character pointer) that doesn't really have anything to do with other stuff in the patch.  I'd recommend splitting optimizations like this into a separate patch.

I'm going to step in here and say that I find the idea of taking a function and moving it to be in-line is a step backwards when it comes to cleaning up code. Even if get_body_by_line is a simple if check, as you say in the description, it helps to have it in a function instead of having a somewhat cryptic if statement in-line.


On June 21, 2012, 3:48 p.m., gareth wrote:
> > I've just taken a brief minute look at this so far, but I'm just going to point out that there are a lot of changes in this patch.  Would you be opposed to breaking this up into a series of smaller patches? There are a number of benefits to this, among them being that if you introduce a problem with these changes that causes a test failure or something we can reproduce later on, we'll be able to pinpoint which change specifically causes this more easily. Another benefit would be that smaller reviews tend to be approved much more quickly.

I'm going to disagree with this. This is already a smaller portion of a larger patch that introduces a couple of new features. Breaking this up further probably isn't necessary. Maybe it's just because the reviews I've been looking at lately have been absolutely massive, but this changeset doesn't really feel all that big to me.


- Mark


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/1993/#review6541
-----------------------------------------------------------


On June 19, 2012, 10:08 p.m., gareth wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/1993/
> -----------------------------------------------------------
> 
> (Updated June 19, 2012, 10:08 p.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Summary
> -------
> 
> This is the first part split out from the much larger patch at https://reviewboard.asterisk.org/r/1976/ - it contains just the various chan_sip code cleanups.
> 
> - struct sip_refer converted to use the stringfields API.
> 
> - sip_{refer|notify}_allocate -> sip_{notify|refer}_alloc to match other *alloc functions.
> 
> - Replace get_msg_text, get_msg_text2 and get_pidf_body -> No, not get_pidf_msg_text_body3 but get_content, to match add_content.
> 
> - get_body doesn't get the request body, renamed to get_content_line.
> 
> - get_body_by_line doesn't get the body line, and is just a simple if test. Moved code inline and removed function.
> 
> - Remove camelCase in struct sip_peer peer state variables, onHold -> onhold, inUse -> inuse, inRinging -> ringing.
> 
> - Remove camelCase in struct sip_request rlPart1 -> rlpart1, rlPart2 -> rlpart2.
> 
> - Rename instances of pvt->randdata to pvt->nonce because that is what it is, no need to update struct sip_pvt because _it already has a nonce field_.
> 
> - Removed struct sip_pvt randdata stringfield.
> 
> - Remove useless (and inconsistent) 'header' suffix on variables in handle_request_subscribe.
> 
> - Use ast_strdupa on Event header in handle_request_subscribe to avoid overly complicated strncmp calls to find the event package.
> 
> - Move get_destination check in handle_request_subscribe to avoid duplicate checking for packages that don't need it.
> 
> - Move extension state callback management in handle_request_subscribe to avoid duplicate checking for packages that don't need it.
> 
> - Remove duplicate append_date prototype.
> 
> - Rename append_date -> add_date to match other add_xxx functions.
> 
> - Added add_expires helper function, removed code that manually added expires header.
> 
> - Remove _header suffix on add_diversion_header (no other header adding functions have this).
> 
> - Don't pass req->debug to request handle_request_XXXXX handlers if req is also being passed.
> 
> - Don't pass req->ignore to check_auth as req is already being passed.
> 
> - Don't create a subscription in handle_request_subscribe if p->expiry == 0.
> 
> - Don't walk of the back of referred_by_name when splitting string in get_refer_info
> 
> - Remove duplicate check for no dialog in handle_incoming when sipmethod == SIP_REFER, handle_request_refer checks for that.
> 
> 
> Diffs
> -----
> 
>   /trunk/channels/chan_sip.c 369065 
>   /trunk/channels/sip/include/sip.h 369065 
>   /trunk/channels/sip/security_events.c 369065 
> 
> Diff: https://reviewboard.asterisk.org/r/1993/diff
> 
> 
> Testing
> -------
> 
> It compiles.
> 
> 
> Thanks,
> 
> gareth
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20120622/4f7b897d/attachment-0001.htm>