3 2

[alpine-devel] webkitfltk segfaults in do_relocs

Isaac Dunham
Details
Message ID
<20141119164412.GB16895@newbook>
Sender timestamp
1416415455
DKIM signature
missing
Download raw message
Hello,
I've been trying to get WebkitFLTK[1] working on Alpine, and I have it
building.
But as usual, that's the easy part. The test program, 
Source/WebKit/fltk/testapp/testapp, segfaults in do_relocs.
I'm attaching the output of bt full.

Build dependencies include:
g++ >= 4.7.3
python2 perl ruby 
icu-dev cairo-dev harfbuzz-dev libxslt-dev sqlite-dev fltk-dev

harfbuzz must be built with ICU, as alpine does; fltk must be 1.3.3
or later; Alpine provides python2 as python.

Build directions are in README.asciidoc, and amount to:
make -C Source/WTF/wtf && \
make -C Source/JavaScriptCore gen && \
make -C Source/JavaScriptCore && \
make -C Source/WebCore && \
make -C Source/WebKit/fltk


Additionally, at least a gig of RAM per thread will be needed,
and a very fast processor or lots of time.
On my laptop with an Atom N270 and a gigabyte of ram, I get 
occasional swap thrashing, and the compilation takes a little over
11 hours.

So if someone who has a faster system can look at this, I'd be
grateful.

Thank you,
Isaac Dunham

[1] https://github.com/clbr/webkitfltk

[alpine-devel] Re: [musl] webkitfltk segfaults in do_relocs

Isaac Dunham
Details
Message ID
<20141119190444.GA17350@newbook>
In-Reply-To
<20141119165050.GK22465@brightrain.aerifal.cx> (view parent)
Sender timestamp
1416423887
DKIM signature
missing
Download raw message
On Wed, Nov 19, 2014 at 11:50:50AM -0500, Rich Felker wrote:
> On Wed, Nov 19, 2014 at 08:44:15AM -0800, Isaac Dunham wrote:
> > Hello,
> > I've been trying to get WebkitFLTK[1] working on Alpine, and I have it
> > building.
> > But as usual, that's the easy part. The test program, 
> > Source/WebKit/fltk/testapp/testapp, segfaults in do_relocs.
> > I'm attaching the output of bt full.
> 
> Could you also provide a disassembly and register dump at the point of
> crash?
> 
Attaching output of gdb disassemble and info all-registers.

It's segfaulting "at line 365 of src/ldso/dynlink.c", for what that's worth.

I'm providing the link command as well in case the problem is there.

Thanks,
Isaac Dunham

[alpine-devel] Re: [musl] webkitfltk segfaults in do_relocs

Rich Felker
Details
Message ID
<20141119165050.GK22465@brightrain.aerifal.cx>
In-Reply-To
<20141119164412.GB16895@newbook> (view parent)
Sender timestamp
1416415850
DKIM signature
missing
Download raw message
On Wed, Nov 19, 2014 at 08:44:15AM -0800, Isaac Dunham wrote:
> Hello,
> I've been trying to get WebkitFLTK[1] working on Alpine, and I have it
> building.
> But as usual, that's the easy part. The test program, 
> Source/WebKit/fltk/testapp/testapp, segfaults in do_relocs.
> I'm attaching the output of bt full.

Could you also provide a disassembly and register dump at the point of
crash?

Rich


---
Unsubscribe:  alpine-devel+unsubscribe@lists.alpinelinux.org
Help:         alpine-devel+help@lists.alpinelinux.org
---

[alpine-devel] Re: webkitfltk segfaults in do_relocs (partly solved)

Isaac Dunham
Details
Message ID
<20141122052430.GB28486@newbook>
In-Reply-To
<20141119164412.GB16895@newbook> (view parent)
Sender timestamp
1416633871
DKIM signature
missing
Download raw message
Hello,
Thanks for all the advice. Now it gets a lot further:
I built webkitfltk with -fno-pie, and the test program loaded, showed 
a window, and crashed somewhere in JavaScriptCore.

I tried running Source/JavaScriptCore/jsc, and it shows a problem that
was previously reported:

$ Source/JavaScriptCore/jsc
Error compiling builtin: 
ASSERTION FAILED: result
parser/Parser.h(970) : WTF::PassRefPtr<T> JSC::parse(JSC::VM*, const JSC::SourceCode&, JSC::FunctionParameters*, const JSC::Identifier&, JSC::JSParserStrictness, JSC::JSParserMode, JSC::ParserError&, JSC::JSTextPosition*) [with ParsedNode = JSC::ProgramNode]
Segmentation fault (core dumped)


Upstream has set thread stack size to 1 megabyte regardless of libc, 
figuring that 8 megabytes is too much.
I'm building with -DUSE_SYSTEM_MALLOC=1.
Looking through the old thread on Webkit javascript crashing, I found an 
excerpt of a diff, but could not figure out where it's supposed to apply
(grep -r didn't find anything pertinent that I could see).

Thanks,
Isaac Dunham



---
Unsubscribe:  alpine-devel+unsubscribe@lists.alpinelinux.org
Help:         alpine-devel+help@lists.alpinelinux.org
---