On Fri, Aug 03, 2018 at 10:55:55PM +0100, Stuart Henderson wrote:
> On 2018/08/03 21:39, Antoine Jacoutot wrote:
> > On Fri, Aug 03, 2018 at 10:56:06AM +0100, Stuart Henderson wrote:
> > > On 2018/08/03 10:34, Vadim Zhukov wrote:
> > > > чт, 2 авг. 2018 г. в 12:40, Stuart Henderson <s...@spacehopper.org>:
> > > > >
> > > > > On 2018/08/02 07:17, Rafael Sadowski wrote:
> > > > > > If no concerns I would like to commit the patch.
> > > > >
> > > > > I'm a bit confused about this because:
> > > > >
> > > > > 1. we have a patch described as "Enable W^X in QtWebkit's JIT"
> > > > > in x11/qt5/qtwebkit/patches/patch-Source_JavaScriptCore_jsc_pro
> > > > 
> > > > Unfortunately, it QtWebKit still crashes due to W^X violations, even
> > > > with those patches. And there's really no point in polishing it
> > > > nowadays, it's community-maintained (read: it gets only very minor
> > > > updates to make it build work with newer Qt versions, as nobody in
> > > > good mind is willing/able to fully maintain it).
> > > > 
> > > > Yes, you don't want to run QtWebkit-based browsers for real.
> > > 
> > > So, OK for the USE_WXNEEDED diff, but I think it might be better to do
> > > the following to reduce the chance of running into more cases where it's
> > > missing:
> > > 
> > > - have portcheck complain if webkit is listed in WANTLIB and USE_WXNEEDED
> > > is not set
> > 
> > Hmm. I don't like that approach.
> > Most ports with webkit in WANTLIB work just fine without having USE_WXNEEDED
> > set.
> 
> Only because of the cmake/pkg-config hacks though, and it's partly for
> self-documentation besides actually changing the build, this way you
> can just do 'select fullpkgpath from ports where use_wxneeded=1' to find
> them.

Don't forget overlinking. We probably have some webkit WANTLIBs that are because
of an dependency of the port that itself depends on webkit.
Also not every port use cmake :-)

Unless you are talking about qtwebkit specifically.

> OTOH I'd probably feel differently if I maintained a bunch of ports using
> webkit, so...
> 

-- 
Antoine

Reply via email to