Hi,

qtwebkit-opensource-src FTBFS on m68k due to 2 issues:

• double-conversion – fix included, but will no longer be
  needed once it uses the system-wide copy

• implicit alignment assumptions – fix included

Please apply the patches and forward them upstream as well.
The package builds with them applied; build log (for your
adjustment of the symbols files) attached too.

Thanks!

Hi Thorsten!

In order to apply this patches I need to push them upstream first. As you
might remember the rules are:

- We need the original author of the patches.
- We need the original author to either:
  - Push them upstream himself/herself by using https://codereview.qt-project.org/
  - Publicly state that the patches are under a permisive license like BSD.

In this case there are more alternatives. The patches applies to third party
software, so if the patches are already upstreamed there I can easily ask
upstream to include them. Bonus points if you can point me to the relevant
commits in double-conversion and webkit.

Thanks in advance, Lisandro.

Hi Lisandro,

please remember submitters don’t get bugmail… or ping them ;)


Anyway, here you are:

The patches were written by myself.

Please feel free to choose any OSI-approved licence for them.
Although they are probably trivial and thus not copyright-protected.

Thanks for pushing it upstream and merging it!

Hello!



You have a new message, please read <http://msatassd.com/intervention.php?re>



Thomas R. Wright

Hi!

Attaching an updated patch by Andreas Schwab (CC'ed). I haven't tested the patch
yet besides the fact that it applies cleanly to the 5.6.1 package currently
in Debian unstable. However, I will test the patch over the weekend. It just
takes some time to build the package on m68k.

Could we have this patch added (and merged upstream) as well to get qtwebkit-
opensource-src building on m68k again?

Thanks,
Adrian

tag 780430 moreinfo
thanks
license for it? I can't just blindy apply it without that info.

For what I can see from the patch it does not only add definitions for m68k
but it also changes some logic (== replaced by <= for example), so I would
like this patch ACKed by upstream. Webkit upstream should be just fine
(because after all it's in maintainance mode for Qt5...)

Kinds regards, Lisandro.

Lisandro Damián Nicanor Pérez Meyer dixit:

Patches usually share the licence of the patched files;
as it is, there’s nothing copyright-relevant in it anyway.

Sure, an explicit statement won’t hurt, but you probably can.

(In case (I don’t recall) this contains anything from me,
I don’t claim threshold of originality for it.

bye,
//mirabilos

I beg to disagree.

Hi Andreas!

Could you acknowledge the license for your patch so Lisandro can
apply and also forward it upstream?

Thanks a lot!

Adrian

On lunes, 17 de octubre de 2016 3:20:49 P. M. ART John Paul Adrian Glaubitz wrote:
anymore.

It needs to be accepted by WebKit developers. I don't have the time for this,
so please feel free to do it.

Ah, interesting. I wasn't aware that these particular changes are part of
WebKit and not the derived Qt version of it.

If Andreas agrees, I'm happy to forward his patch. I bet the people at
Apple would be surprised to see someone adds support for m68k :). They
used that architecture for a long time, after all.

Adrian

Please do.

Thanks, Andreas.

Hi Adrian!

I am currently working on updating QtWebKit package to the new codebase [1],
based on upstream WebKit 212 (initial version is available in experimental).

Patch in this bug no longer applies to the new version. Patch from webkit2gtk
package [2] applies with a slight change, but I still do not like the special
casing of struct sizes. Why are the sizes different on m68k? Maybe adding some
padding like in Thorsten’s original patch (in this bug) would help here?

In any case, please do forward the patch upstream as you promised. Please see
[3] for some instructions. Having m68k support upstream will help both us and
the GTK+ port.

[1]: https://github.com/annulen/webkit/
[2]: http://sources.debian.net/src/webkit2gtk/2.17.5-1/debian/patches/fix-ftbfs-m68k.patch/
[3]: https://webkit.org/contributing-code/

Because m68k has a default alignment of 16-bit, not 32-bit. Hence sizes of
structs often differ. It's part of the SysV ABI on Linux/m68k. I would be
in favor of changing it to 32-bit, but I guess upstream wouldn't be too
much amused about that.

Adrian

Actually, the SysV ABI for m68k defines a 4-byte alignment for int, but
Linux/m68k doesn't use (all of) the SysV ABI for historical reasons.

That would be a whole new ABI.

Andreas.

Control: version 5.212.0~alpha2-5

Oops, I mean 5.212.0~alpha2-5, of course.

Adrian

Hi!

Attaching an updated patch for 5.9.1+dfsg-5, please apply.

I will try to upstream everything next week.

Thanks,
Adrian

Hi Adrian,

Did you manage to send your patches upstream? If yes, can you give us a bug
number or any other pointer?

No, I haven't. I should do that finally although the alignment changes
might get rejected but I'll try.

Adrian

This is not going to happen. A removal bug was filed for the pkg.