Mail Archives: djgpp-workers/1999/12/09/05:10:09
On Wed, 8 Dec 1999, Laurynas Biveinis wrote:
> > Here, you need to include <libc/stubs.h>, so that non-ANSI and
> > non-Posix functions like findfirst and open don't pollute the
> > namespace, like in the fragment above.
> >
> > Also, readlink needs to be added to the stubs list, since ANSI and
> > Posix functions call it, directly or indirectly.
>
> So I added a #define readlink __readlink in <libc/stubs.h> under
> section ``DJGPP functions'' and #include <libc/stubs.h> to the top
> of readlink.c. Is it correct? (I haven't dealed with it before)
Yes, this is correct. You should verify that, when you build the
library after these changes, the build process creates a new stub,
stubNNNN.S, in the src/libc/stubs directory which defines a function
`readlink' ("_readlink" in asm) that simply does a "jmp ___readlink".
> How could I skip such things?
Relax, you did a good job. It's just that the eyes of someone who
sees the sources for the first time aren't used to skip the parts that
are familiar to you, so they catch things that you miss. It happens
to everybody. That's why posting sources here is a good idea.
> > This might be incompatible with the proposed chroot changes, since the
> > root might be simulated. Please put a FIXME comment here, so that
> > this is tested in due time.
>
> What changes? I'm on this mailing list since April or so, but can't
> recall anything.
It was discussed before April, the last message I have is from March.
Search the archives of djgpp-workers for "chroot patches".
Please note that I didn't mean to say that you must make your changes
100% compatible with chroot support, unless you really want to. Since
those changes were not yet integrated into the CVS tree, they might as
well have a slightly different shape when they are checked in. What I
asked for was a comment that would (hopefully) attract someone's
attention that there might be some interaction between these two that
needs to be tested at some point.
- Raw text -