Re: [webkit-dev] Type cast by using toFoo()

2014-01-20 Thread Ravi Kasibhatla
I am the original author of style checker patch reported in previous mail
by Gyuyoung.

My logic for checking the error is simple:
- In the patch, I check for any line which contains regex static_castFoo*.
- If line does contain it, I pick the regex part Foo and search for Foo.h
in the codebase.
- In Foo.h, I check for toFoo signature and if it is present, I ask the
user to use it. If the signature is not present, I throw the message of
adding the toFoo() and using it.
- In cases, where Foo.h is not found, I don't report any error (which I
plan to look in future for more refining).

I have raised the bug for
tracking this change in WebKit and will soon be porting the patch which got
committed in blink (

Ravi Kasibhatla.

On Tue, Oct 1, 2013 at 11:22 AM, Gyuyoung Kim gyuyoung@webkit.orgwrote:

 My plan is to show style error when submitted patch doesn't use toFoo()
 though toFoo exists. This idea was originated from blink commit. However,
 it was reverted because of some regression.

 If my understanding is correct, the toFoo() style checker checks if there
 is toFoo() in a class. If uploaded patch uses static_cast instead of
 toFoo() though there is toFoo(), style checker will generate style error.

 Anyway, I think I need to investigate the commit and consider the idea


 On Tue, Oct 1, 2013 at 3:20 AM, Ryosuke Niwa wrote:

 On Mon, Sep 30, 2013 at 10:52 AM, Yong Li

 Bottom line is turning on RTTI in debug build?

 Style checker analyzes the code statically.  It's nothing to do with
 runtime assertions.  If that wasn't clear enough, style check happens
 before WebKit is ever built.

 - R. Niwa

 webkit-dev mailing list

 webkit-dev mailing list

webkit-dev mailing list

Re: [webkit-dev] WebKit on MIPS: Nix port?

2014-01-20 Thread Filip Pizlo
WebKit is not meant as a playground for experiments. 

Who uses MIPS?

Why hasn't there ever been a bot that runs tests?

Why can't the port be maintained downstream?


 On Oct 8, 2013, at 11:25 PM, Gergely Kis wrote:
 Dear WebKit Developers,
 With the departure of the Qt port, we are now actively looking for a
 primary port for using on MIPS. We are currently considering the Nix
 port, which is currently being merged into WebKit trunk.
 If you are using WebKit on MIPS with a different port in WebKit trunk
 (e.g. GTK or EFL), please contact us.
 Of course, we still support QtWebKit on MIPS outside of WebKit trunk,
 and plan to continue supporting it while QtWebKit is maintained.
 Best Regards,
 webkit-dev mailing list
webkit-dev mailing list

Re: [webkit-dev] How did Apple build libicucore.dylib?

2014-01-20 Thread Mark Rowe
Moving webkit-dev to BCC since this is off-topic for the list.

On Jan 10, 2014, at 13:45, Eric Wing wrote:

 (I tried webkit-help but didn't get any responses, so I thought maybe
 this list might be more appropriate for this question. Sorry for the
 duplicate if otherwise.)
 I am attempting to build JavaScriptCore for my own embedded
 application use. I noticed that Mac and iOS JavaScriptCore have a
 dependency on a library called libicucore.
 When I build ICU myself from the , I get a handful of different icu
 libraries, but nothing named core.
 While I can get a working JavaScriptCore built against my ICU
 libraries (I need icuuc, icui18n, and icudata), my final binary size
 is huge due to the ICU libraries I link in. I tried statically linking
 my icu libraries, but the smallest my binary gets is about 30MB per
 architecture. When I dynamically link ICU, I can tell that the major
 contribution of the bloat is due to the ICU libraries.
 Looking at Apple's icucore binary, it is about 6.5MB for 3
 architectures which is tiny compared to my 3 icu libraries. I'm
 speculating that Apple ripped out all the stuff they don't need and
 merged in all the parts they do need into the single libicucore.
 Can somebody confirm this is what Apple did, and can somebody tell me
 how I can reproduce that? (I did try downloading Apple's 10.9 source
 for ICU from MacOSForge, but it built the normal ICU stuff and not
 libicucore, nor did I notice any special build options.)

Disclaimer: I know very little about ICU, this is just what I found from 
spending a few minutes looking at their source.

If you look at the you 
can see the Makefile used when building ICU. It functions as a wrapper around 
ICU’s regular build process. I think the key option to shrink the library size 
is --with-data-packaging=archive. This puts the ICU data tables in a standalone 
file (/usr/share/icu/icudt51l.dat) rather than embedding them in the library. 
You'll also find the logic for packaging things up in to a library named 
libicucore.dylib here.

- Mark

webkit-dev mailing list

[webkit-dev] Win EWS bots back online

2014-01-20 Thread Roger Fong
Hey all,

Just a heads up.
The Windows EWS bots are back online.
Let me know if you’re still seeing any issues with them.

webkit-dev mailing list