The move from emem to wmem is already breaking compatibility in a
significant way (especially as we remove more and more emem functions
completely), and the current trunk is probably a good place to break
compatibility: it already contains the expert-info API change, and if
the next release is 2.0 (using Qt) then I would prefer to make a clean
break as much as possible. There will be no need to specify the wmem
in wmem_tvb_memdup when eventually there will be no other
tvb_memdup-family functions left.
All that said, I don't have a particularly strong attachment to this
position. If the general consensus ends up being to call them all
wmem_ and keep the existing names as-is, I won't object :)
On Sat, Sep 21, 2013 at 1:57 PM, Pascal Quantin
<pascal.quantin@xxxxxxxxx> wrote:
> Hi Evan,
>
> I started locally converting tvbuff.h functions (I hope to finish the first set tomorrow). For ex tvb_memcpy is now a define for wmem_tvb_memcpy(NULL, ...) and ep_tvb_memcpy is a define for wmem_tvb_memcpy(wmem_packet_scope(), ...) so as to maintain backward compatibility with external dissectors / plugins (I changed also all the function calls in our source tree).
> I just saw the rename you did in r52164. While I understand the purpose I think it is not a move in the right direction as it breaks backward compatibility (and it collides with my changes but it's not the important part ;) ). Would my proposal be an acceptable tradeoff instead?
>
> Cheers,
> Pascal.
>
> Le 21 sept. 2013 à 00:16, Evan Huus <eapache@xxxxxxxxx> a écrit :
>
>> On 2013-09-20, at 5:55 PM, Pascal Quantin <pascal.quantin@xxxxxxxxx> wrote:
>>
>>> Hi all,
>>>
>>> the easy part of the conversion from emem to wmem memory should be almost complete now: dissectors and plugins use the new memory manager (with the exception of uat / initialization routines).
>>>
>>> Next, I was thinking about converting our helper functions found in epan module and the various ep_ and se_ functions.
>>> My idea was to update (and rename!) those functions with the following scheme:
>>> tvb_get_ephemeral_string -> tvb_get_wmem_packet_string
>>> tvb_get_seasonal_string -> tvb_get_wmem_file_string
>>> and adding defines to maintain (temporarily?) backward compatibility:
>>> #define tvb_get_ephemeral_string tvb_get_wmem_packet_string
>>> #define tvb_get_seasonal_string tvb_get_wmem_file_string
>>>
>>> Of course we will discover some misuse of ep/se memory (like for uat) that will require more thinking, but it would be one step forward.
>>
>> I was hoping (at least for functions with both an ep and se version) to make the pool a parameter, so for example
>>
>> tvb_get_string(pool, ...)
>>
>> This can also replace the glib versions of such functions automatically by passing NULL as the pool (wmem provides this "for free").
>>
>> The macros are a good idea.
>>
>>> Thoughts?
>>>
>>> Regards,
>>> Pascal.
>>> ___________________________________________________________________________
>>> Sent via: Wireshark-dev mailing list <wireshark-dev@xxxxxxxxxxxxx>
>>> Archives: http://www.wireshark.org/lists/wireshark-dev
>>> Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev
>>> mailto:wireshark-dev-request@xxxxxxxxxxxxx?subject=unsubscribe
>> ___________________________________________________________________________
>> Sent via: Wireshark-dev mailing list <wireshark-dev@xxxxxxxxxxxxx>
>> Archives: http://www.wireshark.org/lists/wireshark-dev
>> Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev
>> mailto:wireshark-dev-request@xxxxxxxxxxxxx?subject=unsubscribe
> ___________________________________________________________________________
> Sent via: Wireshark-dev mailing list <wireshark-dev@xxxxxxxxxxxxx>
> Archives: http://www.wireshark.org/lists/wireshark-dev
> Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev
> mailto:wireshark-dev-request@xxxxxxxxxxxxx?subject=unsubscribe