Add nacl .globl symbols to framework.order.

Add nacl .globl symbols to framework.order.

TBR=mark@chromium.org, thomasvl@chromium.org


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=33871

[Mark Mentovai, 2009-12-05 07:24:21]

LGTM if this is currently breaking our build as a temporary matter,
but the NaCl people told me in the past that these symbols do not need
to be global.  In fact, I fixed them and made them not-so-global in
the past.

robertm and gregoryd?

Mark

jrg wrote:
> Reviewers: Mark Mentovai, TVL,
>
> Description:
> Add nacl .globl symbols to framework.order.
>
>
> Please review this at http://codereview.chromium.org/464046
>
> SVN Base: svn://chrome-svn/chrome/trunk/src/
>
> Affected files:
>  M     chrome/app/framework.order
>
>
> Index: chrome/app/framework.order
> ===================================================================
> --- chrome/app/framework.order  (revision 33840)
> +++ chrome/app/framework.order  (working copy)
> @@ -54,5 +54,9 @@
>  ___gcov_merge_add
>  ___gcov_fork
>
> +# Written in asm as a .globl.  (Is that necessary?)
> +_NaClSwitch
> +_NaClSyscallSeg
> +
>  # _ChromeMain must be listed last.  That's the whole point of this file.
>  _ChromeMain
>
>
>

[Mark Mentovai, 2009-12-06 16:31:17]

jrg, I'd like to know more about why this was necessary for you.  It looks like
these are still not exported symbols in NaCl.

native_client/src/trusted/service_runtime/arch/x86_32/nacl_switch.S:

.globl  IDENTIFIER(NaClSwitch)
        HIDDEN(NaClSwitch)
IDENTIFIER(NaClSwitch):


native_client/src/trusted/service_runtime/arch/x86_32/nacl_syscall.S:

.globl  IDENTIFIER(NaClSyscallSeg)
        HIDDEN(NaClSyscallSeg)
IDENTIFIER(NaClSyscallSeg):

where HIDDEN is defined in
native_client/src/trusted/service_runtime/nacl_config.h on Mac OS X as:

# define HIDDEN(n)  .private_extern IDENTIFIER(n)

A symbol marked .private_extern should not be showing up as a global symbol in
the final linked product.  jrg, what configuration are you building in that
might have an impact on that?

[John Grabowski, 2009-12-06 17:51:56]

I did a GYP_DEFINES='coverage=1' gclient runhooks --force, then built the
"coverage" target.  When I did, these symbols showed up as part of the
verify_order step.  (I don't recall if I was compiling on 10.5 or 10.6 at
the time; not sure it matters.)

I did not speculatively guess here.

jrg

On Sun, Dec 6, 2009 at 8:31 AM, <mark@chromium.org> wrote:

> jrg, I'd like to know more about why this was necessary for you.  It looks
> like
> these are still not exported symbols in NaCl.
>
> native_client/src/trusted/service_runtime/arch/x86_32/nacl_switch.S:
>
> .globl  IDENTIFIER(NaClSwitch)
>        HIDDEN(NaClSwitch)
> IDENTIFIER(NaClSwitch):
>
>
> native_client/src/trusted/service_runtime/arch/x86_32/nacl_syscall.S:
>
> .globl  IDENTIFIER(NaClSyscallSeg)
>        HIDDEN(NaClSyscallSeg)
> IDENTIFIER(NaClSyscallSeg):
>
> where HIDDEN is defined in
> native_client/src/trusted/service_runtime/nacl_config.h on Mac OS X as:
>
> # define HIDDEN(n)  .private_extern IDENTIFIER(n)
>
> A symbol marked .private_extern should not be showing up as a global symbol
> in
> the final linked product.  jrg, what configuration are you building in that
> might have an impact on that?
>
>
> http://codereview.chromium.org/464046
>

[Mark Mentovai, 2009-12-06 19:47:45]

John Grabowski wrote:
> I did a GYP_DEFINES='coverage=1' gclient runhooks --force, then built the
> "coverage" target.  When I did, these symbols showed up as part of the
> verify_order step.  (I don't recall if I was compiling on 10.5 or 10.6 at
> the time; not sure it matters.)
> I did not speculatively guess here.

I didn't think you did.  I just needed to know what configuration you
were using so I could investigate this further.

Mark