[OpenIndiana-discuss] Tasks to focus on -- firefox

[Carsten Grzemba]

Am 08.01.21 10:25 schrieb "Joshua M. Clulow"  <josh at sysmgr.org>: 
> 
> On Thu, 7 Jan 2021 at 23:34, Carsten Grzemba via openindiana-discuss
> <openindiana-discuss at openindiana.org> wrote:
> > > >On 07.01.21 23:47, Joshua M. Clulow via openindiana-discuss wrote:
> > > >>If you hit any trouble with Rust, I'd love to hear about it! We have
> > > >>a good relationship with the Rust project itself, so we should be able
> > > >>to get things fixed as they arise, without needing to float additional
> > > >>patches.
> > I try to build firefox 78esr for openindiana. The problems with Rust are much smaller than a while ago.
> 
> I took a spin at building Firefox last year, and while I didn't get a
> complete build finished (ran out of time) it was honestly easier to
> get master to build than ESR. The main branch usually uses updated
> crates, and thus tends to have _better_ support for illumos than the
> older ESR branches.
> 
> > But I stopped to use the rustc installed with rustup because this use a toolchain naming like x86_64-unknown-illumos.
> >
> > For that the rustc sources build instruction have to extent for target_os "solaris" to target_os "illumos", which is an additional effort.
> >
> > So I updated the oi-userland build reciept for rust to provide an newer version (1.41.1) where the target tripple x86_64-sun-solaris is used.
> > So there are no patching of lib.rs and others is necessary.
> > Or it is possible with rustup to get the target toolchain x86_64-sun-solaris also?
> 
> The old "sun-solaris" targets are not really being maintained by the
> illumos community at this point -- either in the Rust toolchain
> itself, or in the crates in the Rust ecosystem at large. I would
> recommend using the illumos target and getting any build issues fixed
> upstream in whatever crates have remaining platform-specific issues.
> 
> I and some other folks have been doing that sort of thing a lot
> throughout 2019-2020. Most or all of our patches and testing are for
> the illumos target, even if the Solaris targets sometimes also come
> along for the ride.
> 
> I think it'd be best if OI ships an illumos target toolchain for Rust,
> rather than the increasingly obsolete solaris target. Then we're all
> testing and working on the same bits!
> 
> 
> Cheers.
> 
> -- 
> Joshua M. Clulow
> http://blog.sysmgr.org
> 

Some other findings related buld FF on OI
<signatureafterquotedtext>- update ICU, if we want use system ICU 
</signatureafterquotedtext>
- update mozilla-nss, if we want system-nss
- update rust
- if we use system-sqlite3, in sqlite3 libsqlite-3 2 symbols not exported because these missing in mapfile
these all are already in my repository on github



- readelf (2.34) of our binutil package is unable to read the libxul.so, readelf (2.33) of Solaris 11.4 can it read. updated readelf (2.35.2) reports less errors but fails also.

 

I deceide to use GCC9 because the trouble to compile with clang was bigger, GCC7 stopped with a suspicouse memory allocation error.


Carsten