Port-alpha archive

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

Re: Access to a netbsd/alpha machine



On Sun, Apr 07, 2019 at 04:23:40PM +0200, Martin Husemann wrote:
> On Sun, Apr 07, 2019 at 02:16:33PM +0000, coypu%sdf.org@localhost wrote:
> > On Sun, Apr 07, 2019 at 03:39:56PM +0200, Martin Husemann wrote:
> > > > - If running before 9.0, disable initfini array, otherwise constructors
> > > > won't run.
> > > 
> > > Sorry, I can't parse the "If running before 9.0" - can you explain?
> > > 
> > 
> > I think christos enabled it in the same binutils update,
> > https://v4.freshbsd.org/search?q=initfini&committer%5B%5D=christos&sort=commit_date
> > 
> > (I was naively thinking "pick the same version as -current, that one is
> > known to work on netbsd/alpha" :))
> 
> Sorry, still completely unable to parse any of this.
> 
> What "it"? What does "if running before 9.0" mean?
> 
> 
> Martin

As I understand it,
running constructor and destructors is the responsibility of the runtime
link editor. there is more than one way to specify it.
the shiny way to do it is with "initfini array".

if you run old binaries, you need the existing way anyway, so not much
benefit to switch.

but since things like the latest binutils will default to assuming
initfini array, maybe we should just support initfini array way of
specifying constructors and destructors, and not fight uphill battles.

christos decided to make all netbsd architectures support initfini
array.

but he did this in netbsd-current. the build machine I used was 8.0
(thanks jklos) and it does not contain this change yet.

if you are running 8.99.x or anything that comes after it like 9.0 then
it will probably have support for it.
I'm future proofing the post by referring to 9.0 :)


Home | Main Index | Thread Index | Old Index