| 1 |
On 09/17/14 15:41, Markos Chandras wrote: |
| 2 |
> -----BEGIN PGP SIGNED MESSAGE----- |
| 3 |
> Hash: SHA512 |
| 4 |
> |
| 5 |
> On 09/17/2014 08:13 PM, Anthony G. Basile wrote: |
| 6 |
>> On 09/17/14 13:50, Markos Chandras wrote: On 09/17/2014 02:38 PM, |
| 7 |
>> Ian Stakenvicius wrote: |
| 8 |
>>>>> On 17/09/14 04:31 AM, Micha³ Górny wrote: |
| 9 |
>>>>>> Dnia 2014-09-13, o godz. 10:47:49 Markos Chandras |
| 10 |
>>>>>> <hwoarang@g.o> napisa³(a): |
| 11 |
>>>>>>> Here is some weirdness with eg mips64/n32 multilib |
| 12 |
>>>>>>> profile when trying a world update |
| 13 |
>>>>>>> |
| 14 |
>>>>>>> [ebuild U ] sys-devel/libtool-2.4.2-r1:2 [2.4.2:2] |
| 15 |
>>>>>>> USE="-static-libs {-test} -vanilla" ABI_MIPS="(n32%*) |
| 16 |
>>>>>>> o32%* -n64%" 0 kB |
| 17 |
>>>>>>> |
| 18 |
>>>>>>> As you can see n32 and o32 are enabled but n64 is not. |
| 19 |
>>>>>>> Obviously this is not full mips64 multilib. This is |
| 20 |
>>>>>>> probably due the portage profile stacking/inheritance |
| 21 |
>>>>>>> problems on mips64, where the mips64/multilib profiles |
| 22 |
>>>>>>> inherit the default o32 one. Michal (multilib CC'd) can |
| 23 |
>>>>>>> provide more information on what exactly goes wrong since |
| 24 |
>>>>>>> he understands the problem better than me. Michal also |
| 25 |
>>>>>>> said that on amd64, the multilib profiles defaults to |
| 26 |
>>>>>>> 64-bit only. I believe this contradicts with what someone |
| 27 |
>>>>>>> expects from MIPS64 where all three ABIs need to be |
| 28 |
>>>>>>> present *by default* unless you override the ABI_MIPS |
| 29 |
>>>>>>> variable in make.conf. Correct? |
| 30 |
>>>>>> Well, long story short we inherit from 'top-level' profile |
| 31 |
>>>>>> that has some o32 settings inside. I believe that it could |
| 32 |
>>>>>> be saner to move those from arch/mips/mips64 -> |
| 33 |
>>>>>> arch/mips/mips64/o32 (like we have /n32 and /n64 there), so |
| 34 |
>>>>>> that instead of having to unset them, we'd just have them |
| 35 |
>>>>>> set for the relevant real profiles. |
| 36 |
>>>>>> However, I'm not sure if this doesn't come with some |
| 37 |
>>>>>> pitfalls. |
| 38 |
>>>>> |
| 39 |
>>>>> Blueness and I talked about this (proper n32 / n64 / o32 |
| 40 |
>>>>> defaults and forces/masks) in #gentoo-dev two or three weeks |
| 41 |
>>>>> ago; I thought we worked out the correct modifications to |
| 42 |
>>>>> profiles to get it right and he had already pushed the |
| 43 |
>>>>> fixes... ?? |
| 44 |
>> I can't see anything. Did you actually push them? What was decided |
| 45 |
>> as the plan for action? |
| 46 |
>> |
| 47 |
>>> I did not have time to get to them. I was going to play with |
| 48 |
>>> two different approaches. One is simply turning off o32 at |
| 49 |
>>> multilib/../n32 level, the other was to restructure the profiles |
| 50 |
>>> entirely and put o32, n32 and 64 on par, not inherit from a hire |
| 51 |
>>> level "default" profile. I'm leaning towards that approach, but |
| 52 |
>>> I'm worried it might play havoc on our users. |
| 53 |
> Well i need to see the structure for your second approach to |
| 54 |
> understand what you have in mind. Perhaps the first option is the |
| 55 |
> safest and ensure some backwards compatibility with existing mips32 users? |
| 56 |
|
| 57 |
Yes it would be more backwards compat, but it is not semetrical. |
| 58 |
|
| 59 |
The regular (ie glibc) profiles would become: |
| 60 |
|
| 61 |
[1] default/linux/mips/13.0/o32 <-change |
| 62 |
[2] default/linux/mips/13.0/n32 |
| 63 |
[3] default/linux/mips/13.0/n64 |
| 64 |
[4] default/linux/mips/13.0/multilib/o32 <-change |
| 65 |
[5] default/linux/mips/13.0/multilib/n32 |
| 66 |
[6] default/linux/mips/13.0/multilib/n64 |
| 67 |
[7] default/linux/mips/13.0/mipsel/o32 <- change |
| 68 |
[8] default/linux/mips/13.0/mipsel/n32 |
| 69 |
[9] default/linux/mips/13.0/mipsel/n64 |
| 70 |
[10] default/linux/mips/13.0/mipsel/multilib/o32 <-change |
| 71 |
[11] default/linux/mips/13.0/mipsel/multilib/n32 |
| 72 |
[12] default/linux/mips/13.0/mipsel/multilib/n64 |
| 73 |
|
| 74 |
The embedded profiles would remain the same since they are o32 only |
| 75 |
(right now): |
| 76 |
|
| 77 |
[13] hardened/linux/musl/mips |
| 78 |
[14] hardened/linux/musl/mips/mipsel |
| 79 |
[15] default/linux/uclibc/mips |
| 80 |
[16] hardened/linux/uclibc/mips |
| 81 |
[17] default/linux/uclibc/mips/mipsel |
| 82 |
[18] hardened/linux/uclibc/mips/mipsel |
| 83 |
|
| 84 |
I doubt we would ever do a mutlilib musl or uclibc. But we might have a |
| 85 |
uclibc n32 or n64, in which case we'd repeat the same as above. |
| 86 |
|
| 87 |
> |
| 88 |
> - -- |
| 89 |
> Regards, |
| 90 |
> Markos Chandras |
| 91 |
> -----BEGIN PGP SIGNATURE----- |
| 92 |
> Version: GnuPG v2 |
| 93 |
> |
| 94 |
> iQF8BAEBCgBmBQJUGeP/XxSAAAAAAC4AKGlzc3Vlci1mcHJAbm90YXRpb25zLm9w |
| 95 |
> ZW5wZ3AuZmlmdGhob3JzZW1hbi5uZXRGRDlGMzA4MUI2MzBDODQ4RDBGOEYxMjQx |
| 96 |
> RjEwRUQ0QjgxREVCRjE5AAoJEB8Q7UuB3r8Z5msH/1YBYb9trZYGrkXkR0FIcx9Y |
| 97 |
> ltJ8PL7jGG1B4NO0PJJpwehMSsxFbkpq3VT9ahrVq4+K58/3XRDmoWGEsWpBIo3o |
| 98 |
> KHCmCOoC2KPmGFEXofKsD7iAlb83X5/KsHVhHioUy/5D7JMf4PrPLPjkMtRFU0oR |
| 99 |
> h9+hah+3tSMO5QUh4blXnYJ4LeE298GjPJMwPtMlhx4uRUyXeRhUfuINzdf9uMBV |
| 100 |
> FFHfJKOfsr9aizUpJxza/Wph+IA+NVGTCekZzWQ6gSZW+MxiF3mAeLFj6I6g+0lI |
| 101 |
> Ga8+Z1Bs/JM7P4csLJ2Sp+ccgaIGXg6xU/BtlpKyJl745mpBAt58w17762+ivjs= |
| 102 |
> =PqPy |
| 103 |
> -----END PGP SIGNATURE----- |
| 104 |
> |
| 105 |
|
| 106 |
|
| 107 |
-- |
| 108 |
Anthony G. Basile, Ph.D. |
| 109 |
Gentoo Linux Developer [Hardened] |
| 110 |
E-Mail : blueness@g.o |
| 111 |
GnuPG FP : 1FED FAD9 D82C 52A5 3BAB DC79 9384 FA6E F52D 4BBA |
| 112 |
GnuPG ID : F52D4BBA |
Archives