Re: static link csh/tcsh dumps core

[christos%zoulas.com@localhost (Christos Zoulas)]

On Mar 11,  5:51pm, martin%duskware.de@localhost (Martin Husemann) wrote:
-- Subject: Re: static link csh/tcsh dumps core

| I think we hit a corner case of section "3.2.3 Link-Time Optimization"
| from the referenced paper.
| 
| In this case (static linking) caller knows that the callee shares the
| same GP value, so relies on that being present on return from the setjmp.
| 
| We do not need to save/restore the GP value of the setjmp, but must make
| sure that longjmp() fills the ucontext GP value with its own GP before
| calling setcontext().
| 
| Does this make sense?

It makes sense. I am just wondering what the the pros and cons doing it this
way.

christos