[webkit-dev] Proposal: Remove ENABLE(MATHML)

[Ryosuke Niwa]

On Tue, Oct 24, 2017 at 2:35 AM, Olivier Blin <olivier.blin at softathome.com>
wrote:

> Le 24/10/2017 à 09:19, Frédéric WANG a écrit :
>
>> On 23/10/2017 18:25, Adrian Perez de Castro wrote:
>> > From your description it sounds that what takes more disk space in
>> MathML are
>> > the needed math fonts, and that the amount of code added to WebKit when
>> built
>> > is in the order of hundreds of kilobytes, at most (BTW, do you have
>> figures of
>> > MathML enabld vs. disabled to confirm?). That being the case, it seems
>> fine to
>> > me that MathML support is always enabled.
>> >
>> > Usually my main concern with removing compilation guards is forcing
>> additional
>> > usage of megabytes of disk and/or memory in embedded platforms (which
>> is, for
>> > example, the case for ENABLE_VIDEO: for the GTK+ and WPE ports it
>> completely
>> > avoids the GStreamer dependency, which saves tens of megabytes!). But
>> keeping
>> > the guards to save just a few hundreds of kilobytes is certainly not
>> worth it,
>> > and in this case embedders can choose to not ship the math fonts.
>> >
>> > TL;DR: Fine by me, as it seems the compile flag only saves a few
>> kilobytes in
>> > the built binaries, and the math fonts can't be skipped anyway by
>> embedders.
>>
>> Thanks Adrián. No, I wanted to do it during the Web Engines Hackfest but
>> did not have time to try. Also, Olivier mentioned he would check the
>> difference and tell whether it is significant or not for his project,
>> but I have not heard back from him yet.
>>
>
> Hi,
>
> The difference is about 250 kB on libWPEWebKit.so.0 for a WPE build on my
> ARMv7 target:
>
> - without MathML : 37474160
> - with MathML: 37726056
>
> This can still be significant on low-end devices, with 128 MB of flash
> storage.


That's only 0.6%. I'm pretty certain there are better things to work on if
your focus was to reduce the binary size.
e.g. you could de-inline a bunch of functions in the binding code and
significantly reduce the binary size with a slight perf overhead.

- R. Niwa
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-dev/attachments/20171024/cac0a01b/attachment.html>