* [Buildroot] --with-abi=<empty> ... during gcc build
@ 2007-09-29 17:38 Ulf Samuelsson
2007-09-29 18:28 ` Bernhard Fischer
0 siblings, 1 reply; 6+ messages in thread
From: Ulf Samuelsson @ 2007-09-29 17:38 UTC (permalink / raw)
To: buildroot
gcc configure is passed $(GCC_WITH_ABI)
which is defined as:
toolchain/gcc/Makefile.in:GCC_WITH_ABI:=--with-abi=$(BR2_GCC_TARGET_ABI)
target/Config.in.arch defines BR2_GCC_TARGET_ABI
config BR2_GCC_TARGET_ABI
string
default apcs-gnu if BR2_arm_dunno
default aapcs-linux if BR2_arm_dunno
default iwmmxt if BR2_iwmmxt
default 32 if BR2_mipsel && BR2_OABI
default n32 if BR2_mipsel && BR2_EABI
default o64 if BR2_mips && BR2_OABI
default 64 if BR2_mips && BR2_ABI64
default eabi if BR2_mips && BR2_EABI
default mmixware if BR2_mmix && BR2_ABI_native
default gnu if BR2_mmix && !BR2_ABI_native
default altivec if BR2_powerpc && BR2_ABI_altivec
but there is no definition of BR2_OABI or BR2_EABI.
Suggest the following patch
Index: target/Config.in.arch
===================================================================
--- target/Config.in.arch (revision 20117)
+++ target/Config.in.arch (arbetskopia)
@@ -117,8 +117,10 @@
config BR2_ARM_OABI
bool "OABI"
+ select BR2_OABI
config BR2_ARM_EABI
bool "EABI"
+ select BR2_EABI
endchoice
choice
@@ -635,6 +637,12 @@
default z990 if BR2_s390_z990
default z9-109 if BR2_s390_z9_109
+config BR2_EABI
+ bool
+
+config BR2_OABI
+ bool
+
config BR2_GCC_TARGET_ABI
string
default apcs-gnu if BR2_arm_dunno
@@ -643,10 +651,12 @@
default aapcs-linux if BR2_arm_dunno
default iwmmxt if BR2_iwmmxt
default 32 if BR2_mipsel && BR2_OABI
+ default 32 if BR2_arm && BR2_OABI
default n32 if BR2_mipsel && BR2_EABI
default o64 if BR2_mips && BR2_OABI
default 64 if BR2_mips && BR2_ABI64
default eabi if BR2_mips && BR2_EABI
+ default eabi if BR2_arm && BR2_EABI
default mmixware if BR2_mmix && BR2_ABI_native
default gnu if BR2_mmix && !BR2_ABI_native
default altivec if BR2_powerpc && BR2_ABI_altivec
-----------
It is not complete, since the other ABIs are not defined as well...
Best Regards
Ulf Samuelsson
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Buildroot] --with-abi=<empty> ... during gcc build
2007-09-29 17:38 [Buildroot] --with-abi=<empty> ... during gcc build Ulf Samuelsson
@ 2007-09-29 18:28 ` Bernhard Fischer
2007-09-29 20:32 ` Ulf Samuelsson
0 siblings, 1 reply; 6+ messages in thread
From: Bernhard Fischer @ 2007-09-29 18:28 UTC (permalink / raw)
To: buildroot
On Sat, Sep 29, 2007 at 07:38:50PM +0200, Ulf Samuelsson wrote:
>gcc configure is passed $(GCC_WITH_ABI)
>which is defined as:
>toolchain/gcc/Makefile.in:GCC_WITH_ABI:=--with-abi=$(BR2_GCC_TARGET_ABI)
Well, if it is empty it will not be set, so there is something wrong at
your end, really.
ifneq ($(strip $(subst ",,$(BR2_GCC_TARGET_ABI))),)
# ")))
GCC_WITH_ABI:=--with-abi=$(BR2_GCC_TARGET_ABI)
endif
>
>target/Config.in.arch defines BR2_GCC_TARGET_ABI
>config BR2_GCC_TARGET_ABI
> string
> default apcs-gnu if BR2_arm_dunno
> default aapcs-linux if BR2_arm_dunno
> default iwmmxt if BR2_iwmmxt
> default 32 if BR2_mipsel && BR2_OABI
> default n32 if BR2_mipsel && BR2_EABI
> default o64 if BR2_mips && BR2_OABI
> default 64 if BR2_mips && BR2_ABI64
> default eabi if BR2_mips && BR2_EABI
> default mmixware if BR2_mmix && BR2_ABI_native
> default gnu if BR2_mmix && !BR2_ABI_native
> default altivec if BR2_powerpc && BR2_ABI_altivec
>
>
>but there is no definition of BR2_OABI or BR2_EABI.
>
>Suggest the following patch
Heh. I'm extending the ABI selection for MIPS as we speak. I'll
take care of this.
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Buildroot] --with-abi=<empty> ... during gcc build
2007-09-29 18:28 ` Bernhard Fischer
@ 2007-09-29 20:32 ` Ulf Samuelsson
2007-09-29 20:49 ` Bernhard Fischer
0 siblings, 1 reply; 6+ messages in thread
From: Ulf Samuelsson @ 2007-09-29 20:32 UTC (permalink / raw)
To: buildroot
l?r 2007-09-29 klockan 20:28 +0200 skrev Bernhard Fischer:
> On Sat, Sep 29, 2007 at 07:38:50PM +0200, Ulf Samuelsson wrote:
> >gcc configure is passed $(GCC_WITH_ABI)
> >which is defined as:
> >toolchain/gcc/Makefile.in:GCC_WITH_ABI:=--with-abi=$(BR2_GCC_TARGET_ABI)
>
> Well, if it is empty it will not be set, so there is something wrong at
> your end, really.
> ifneq ($(strip $(subst ",,$(BR2_GCC_TARGET_ABI))),)
> # ")))
> GCC_WITH_ABI:=--with-abi=$(BR2_GCC_TARGET_ABI)
> endif
>
> >
> >target/Config.in.arch defines BR2_GCC_TARGET_ABI
> >config BR2_GCC_TARGET_ABI
> > string
> > default apcs-gnu if BR2_arm_dunno
> > default aapcs-linux if BR2_arm_dunno
> > default iwmmxt if BR2_iwmmxt
> > default 32 if BR2_mipsel && BR2_OABI
> > default n32 if BR2_mipsel && BR2_EABI
> > default o64 if BR2_mips && BR2_OABI
> > default 64 if BR2_mips && BR2_ABI64
> > default eabi if BR2_mips && BR2_EABI
> > default mmixware if BR2_mmix && BR2_ABI_native
> > default gnu if BR2_mmix && !BR2_ABI_native
> > default altivec if BR2_powerpc && BR2_ABI_altivec
> >
> >
> >but there is no definition of BR2_OABI or BR2_EABI.
> >
> >Suggest the following patch
>
> Heh. I'm extending the ABI selection for MIPS as we speak. I'll
> take care of this.
I think this is why I get an EABI compiler
even when OABI is selected in buildroot.
I fo not know if --with-abi=32 is the right thing to do for ARM
when OABI is wanted, so if you know it is different, pls change.
BR
Ulf Samuelsson
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Buildroot] --with-abi=<empty> ... during gcc build
2007-09-29 20:32 ` Ulf Samuelsson
@ 2007-09-29 20:49 ` Bernhard Fischer
2007-09-29 22:12 ` Bernhard Fischer
2007-09-30 15:43 ` Ulf Samuelsson
0 siblings, 2 replies; 6+ messages in thread
From: Bernhard Fischer @ 2007-09-29 20:49 UTC (permalink / raw)
To: buildroot
On Sat, Sep 29, 2007 at 10:32:26PM +0200, Ulf Samuelsson wrote:
>l?r 2007-09-29 klockan 20:28 +0200 skrev Bernhard Fischer:
>> On Sat, Sep 29, 2007 at 07:38:50PM +0200, Ulf Samuelsson wrote:
>> >gcc configure is passed $(GCC_WITH_ABI)
>> >which is defined as:
>> >toolchain/gcc/Makefile.in:GCC_WITH_ABI:=--with-abi=$(BR2_GCC_TARGET_ABI)
>>
>> Well, if it is empty it will not be set, so there is something wrong at
>> your end, really.
>> ifneq ($(strip $(subst ",,$(BR2_GCC_TARGET_ABI))),)
>> # ")))
>> GCC_WITH_ABI:=--with-abi=$(BR2_GCC_TARGET_ABI)
>> endif
Note that this has nothing to do with that. I have a suspicion (comment
the EABI selection of the device/Arm/Config.in locally for now.
Locally!)
>I think this is why I get an EABI compiler
>even when OABI is selected in buildroot.
grep -i abi .config
once again, please.
>
>I fo not know if --with-abi=32 is the right thing to do for ARM
>when OABI is wanted, so if you know it is different, pls change.
I took that unreflected and applied it. Oh well..
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Buildroot] --with-abi=<empty> ... during gcc build
2007-09-29 20:49 ` Bernhard Fischer
@ 2007-09-29 22:12 ` Bernhard Fischer
2007-09-30 15:43 ` Ulf Samuelsson
1 sibling, 0 replies; 6+ messages in thread
From: Bernhard Fischer @ 2007-09-29 22:12 UTC (permalink / raw)
To: buildroot
On Sat, Sep 29, 2007 at 10:49:14PM +0200, Bernhard Fischer wrote:
>On Sat, Sep 29, 2007 at 10:32:26PM +0200, Ulf Samuelsson wrote:
>>I think this is why I get an EABI compiler
>>even when OABI is selected in buildroot.
>
>grep -i abi .config
>once again, please.
>>
>>I fo not know if --with-abi=32 is the right thing to do for ARM
>>when OABI is wanted, so if you know it is different, pls change.
I've reverted this and the renaming by now, and for arm and mips the abi
is fine again (talking about r20130). Please show me your abi settings
like above.
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Buildroot] --with-abi=<empty> ... during gcc build
2007-09-29 20:49 ` Bernhard Fischer
2007-09-29 22:12 ` Bernhard Fischer
@ 2007-09-30 15:43 ` Ulf Samuelsson
1 sibling, 0 replies; 6+ messages in thread
From: Ulf Samuelsson @ 2007-09-30 15:43 UTC (permalink / raw)
To: buildroot
l?r 2007-09-29 klockan 22:49 +0200 skrev Bernhard Fischer:
> On Sat, Sep 29, 2007 at 10:32:26PM +0200, Ulf Samuelsson wrote:
> >l?r 2007-09-29 klockan 20:28 +0200 skrev Bernhard Fischer:
> >> On Sat, Sep 29, 2007 at 07:38:50PM +0200, Ulf Samuelsson wrote:
> >> >gcc configure is passed $(GCC_WITH_ABI)
> >> >which is defined as:
> >> >toolchain/gcc/Makefile.in:GCC_WITH_ABI:=--with-abi=$(BR2_GCC_TARGET_ABI)
> >>
> >> Well, if it is empty it will not be set, so there is something wrong at
> >> your end, really.
> >> ifneq ($(strip $(subst ",,$(BR2_GCC_TARGET_ABI))),)
> >> # ")))
> >> GCC_WITH_ABI:=--with-abi=$(BR2_GCC_TARGET_ABI)
> >> endif
>
> Note that this has nothing to do with that. I have a suspicion (comment
> the EABI selection of the device/Arm/Config.in locally for now.
> Locally!)
>
> >I think this is why I get an EABI compiler
> >even when OABI is selected in buildroot.
>
> grep -i abi .config
> once again, please.
> >
Only OABI set, not EABI.
> >I fo not know if --with-abi=32 is the right thing to do for ARM
> >when OABI is wanted, so if you know it is different, pls change.
>
> I took that unreflected and applied it. Oh well..
BR
Ulf Samuelsson
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2007-09-30 15:43 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-09-29 17:38 [Buildroot] --with-abi=<empty> ... during gcc build Ulf Samuelsson
2007-09-29 18:28 ` Bernhard Fischer
2007-09-29 20:32 ` Ulf Samuelsson
2007-09-29 20:49 ` Bernhard Fischer
2007-09-29 22:12 ` Bernhard Fischer
2007-09-30 15:43 ` Ulf Samuelsson
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox