Re: [Arm] Stack pointer setup thumb
Alexander Fedotov <[email protected]>
| Newsgroups | gmane.comp.lib.newlib |
|---|---|
| Message-ID | <CAN8C2Cr7+AVxpWGBpiN5mWCftQPBfcdxyo8HKGOg98pkESQ2ww@mail.gmail.com> |
On Thu, 7 Mar 2019 at 18:45, Nick Clifton <[email protected]> wrote: > Hi Alex, > > > I hope you are have some memories on this fix :) > > > https://sourceware.org/git/gitweb.cgi?p=newlib-cygwin.git;a=commit;h=026f9d290f5fe85fe9bd3a56be378fc794e28473 > > Not really :-( > > > Is this still relevant ? > > The whole patch, or just part of it ? That part below (with sp initialization) > > > > The problem is that users complaining that it has different behavior > > in ARM and Thumb modes for one Cortex-R core. And they do expected > > that library will make some minimal initialization in both ISA > > versions. > > Well to be honest I really do not remember what prompted this patch in > the first place. So if you have a fix which makes the users happy and > which does not break anything then I would say "go for it". :-) > > > Should we just remove this ? > > +#ifdef __thumb2__ > > + /* XXX Fill in stack assignments for interrupt modes. */ > > +#else > > Sure - I would have no problems with that. The comment was more > of a suggestion that someone needs to make sure that the stack > pointer is set up before interrupts happen, but it does not have > to be in this particular piece of code. Plus presumably this has > already been done, or else many more people would be complaining > about interrupts not working in Thumb mode. > Yep. Basic idea is to have sp initialization for exceptions as well. Actually I already proposed a patch for this piece of code where I moved initialuzation in weak subroutine. But I saved this ifdef there as well. I will update patch then :) > > Cheers > Nick > Thsnk you Nick ! -- Best regards, AF