[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#839048: js24 crashes on ARM64



Sorry for making people confused. The patch I uploaded for mozj24 and mozjs1.8.5 are different.
mozjs24 -- The patch is the back port version of the upstream one. It doesn't change the ABI.
mozjs1.8.5 -- The patch is different with the upstream one. Because in the old js, there might be objects pass from C allocation, only modify the heap doesn't work. The patch changes the tag bits. And ABI is changed also. So it will require to rebuild all the packages which depends on mozjs1.8.5

On 11 December 2016 at 11:08, Ben Hutchings <ben@decadent.org.uk> wrote:
On Sun, 2016-12-11 at 03:24 +0100, John Paul Adrian Glaubitz wrote:
> On 12/11/2016 03:13 AM, Ben Hutchings wrote:
> > Looking at the patch that was added in -4: WTF?  It changes the library
> > ABI, but you didn't bump the soname.  Of course that breaks clients!
>
> That's odd. Why does it break the ABI though? From what I can see, the
> upstream change just forces the JS memory allocator to avoid certain
> regions in memory [1].

That's not the one that got backported though.

> > libmozjs185 appears to have been broken in the same way.
>
> Is it? As far as I know, the patch was not merged to src:mozjs.

You called the patch Modify_tagged_pointer_structure_js1.8.5.patch

In mozjs24 it's called Manually_mmap_on_arm64.patch but it's not what
the name says, it's changing the number of tag bits.

Ben.

> > At this point in the release cycle no more library transitions are
> > allowed, so libmozjs has to be fixed with a kluge like the patch that
> > Zheng Xu sent, not the upstream change.
>
> I agree. But I also think that the ABI break was not intentional.
>
> Adrian
>
> > [1] https://hg.mozilla.org/mozilla-central/rev/dfaafbaaa291
--
Ben Hutchings
Any sufficiently advanced bug is indistinguishable from a feature.


Reply to: