This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc 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: [PATCH] split assume pipe2/dup3/sock_cloexec knobs


> > I wonder if there is something analogous to check-localplt that we
> > could come up with for global-variable GOT slots and root this out
> > generally.
> 
> annoyingly, `readelf -r` doesn't decode the symbols for R_X86_64_RELATIVE 
> relocs.  so i think we might have to fix that first.

I've been noodling with scripts and there's ways to deduce that stuff
without really changing the tools.  (It's certainly easier for RELA
machines than for REL machines.)  I'll noodle some more next week.
(I might wind up just writing a C program using libelf.)

> > More radically, we could conceivably one day just
> > compile everything with -fvisibility=hidden and specifically
> > annotate (via some new libc-symbol.h macro shorthand) each function
> > intended to be exported with an overriding attribute.  Hmm.
> 
> that might just be crazy enough to work ;)

It might, and it might be the more useful discipline to impose on ourselves
(the discipline to annotate exported functions and variables at their
definition sites vs the discipline to annotate everything else at all its
declaration sites).  But I'm sure that what I'll do is write the testing
stuff first anyway, as something to analyze the effects of experimenting
with visibility hacks.


Thanks,
Roland


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