Mailing List Archive

gcc-3.4.4-r1 and kernel-2.4.30
Re: gcc-3.4.4-r1 and kernel-2.4.30 [ In reply to ]
On Dec 4, 2005, at 12:31 AM, mathewt@lucent.com wrote:

> Hello All,
>
> I switched to gcc-3.4.4-r1, rebuilt the tool-chain and was trying
> to recompile the kernel (2.4.30)
>
> gcc would always fail on arch/alpha/kernel/pci_iommu.c
> the error message is
>
> pci_iommu.c: In function `sg_fill':
> pci_iommu.c:510: sorry, unimplemented: inlining failed in call to
> 'sg_fill': recursive inlining
> pci_iommu.c:558: sorry, unimplemented: called from here
> make[1]: *** [pci_iommu.o] Error 1
> make: *** [_dir_arch/alpha/kernel] Error 2
> make: *** Waiting for unfinished jobs....
>
> gcc-3.3.2 seemed to handle this fine. has anyone hit the same issue
> and is there a patch for this ?
>

I had the same problems, I compiled everything with gcc-2.96 on a
RedHat...

> thanks, Mathew
>
>
> --
> gentoo-alpha@gentoo.org mailing list
>

----
dawe

http://dawe.ilbello.com


--
gentoo-alpha@gentoo.org mailing list
Re: gcc-3.4.4-r1 and kernel-2.4.30 [ In reply to ]
Davide Cittaro wrote:
>
> On Dec 4, 2005, at 12:31 AM, mathewt@lucent.com wrote:
>
>> Hello All,
>>
>> I switched to gcc-3.4.4-r1, rebuilt the tool-chain and was trying
>> to recompile the kernel (2.4.30)
>>
>> gcc would always fail on arch/alpha/kernel/pci_iommu.c
>> the error message is
>>
>> pci_iommu.c: In function `sg_fill':
>> pci_iommu.c:510: sorry, unimplemented: inlining failed in call to
>> 'sg_fill': recursive inlining
>> pci_iommu.c:558: sorry, unimplemented: called from here
>> make[1]: *** [pci_iommu.o] Error 1
>> make: *** [_dir_arch/alpha/kernel] Error 2
>> make: *** Waiting for unfinished jobs....
>>
>> gcc-3.3.2 seemed to handle this fine. has anyone hit the same issue
>> and is there a patch for this ?
>>
>
> I had the same problems,

Confirmed. The problem appears with gcc-3.4.4-r1 (stable branch) and
both, vanilla-sources 2.4.30 (stable) and 2.4.32 (testing).

I compiled everything with gcc-2.96 on a
> RedHat...

Problem:
I think the problem is the way in which gcc-3.4.4 manage recursive
inlining. sg_fill is an inline function which call itself and 3.4.4
doesn't like that.

Status:
We have our gcc ninjas working on it, hope to have it fixed soon.

Workaround:
I think is only problem of gcc-3.4.4. I can compile it with gcc-3.3.2
just fine. Since gcc is installed slotted, probably everyone who updated
the gcc compiler can change it using gcc-config and compile the kernel
with an old gcc (if needed).

Thanks guys for reporting.

----------------------------
YosWinK @ gentoo .org
Gentoo Alpha
--
gentoo-alpha@gentoo.org mailing list