This is the mail archive of the mailing list for the newlib project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: ftruncate or chsize in newlib/libgloss?

On Wednesday 26 June 2013 13:03:51 Steve Ellcey wrote:
> I made a configure change to GCC recently so that I could build a
> 'bare-metal' GNU Fortran tool chain that uses newlib and libgloss.
> I want to be able to test this configuration with the GNU simulator
> but I currently get some test failures because newlib/libgloss does not
> implement ftruncate or chsize (at least not for the MIPS system).  I
> think the fix for these failures will be to either implement the needed
> fortran functionality without using ftruncate or chsize or to just XFAIL
> the failing tests in the Fortran testsuite but I wanted to check here
> first to see if there is any interest in implementing either of these
> functions in libgloss or alternatively if there are good reasons why
> we don't want either of these functions in libgloss.
> I see that libgloss/syscall.h has:
> * These should be as small as possible to allow a port to use a trap type
>    instruction, which the system call # as the trap (the d10v for instance
>    supports traps 0..31).
> Which is why I think we may not want ftruncate or chsize in libgloss
> but I want to verify if this is still the case.

maybe we should mark the low space for "important" syscalls only.  for all 
others, let's add them at a higher base.  personally, i wouldn't mind seeing 
more syscalls being supported in libgloss/sim.

Attachment: signature.asc
Description: This is a digitally signed message part.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]