Re: [PATCH v2] build: specify minimum versions of make and binutils

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 28/01/2016 13:47, Jan Beulich wrote:
>>>> On 28.01.16 at 14:02, <ian.campbell@citrix.com> wrote:
>> On Thu, 2016-01-28 at 05:49 -0700, Jan Beulich wrote:
>>>>>> On 28.01.16 at 00:12, <cardoe@cardoe.com> wrote:
>>>> To help people avoid having to figure out what versions of make and
>>>> binutils need to be supported document them explicitly. The version of
>>>> binutils that had to be supported was mentioned in
>>>> http://lists.xenproject.org/archives/html/xen-devel/2016-01/msg00609.ht 
>>>> ml 
>>>> as 2.17 recently. It was decided that the versions should instead be
>>>> GNU binutils 2.16.1 and GNU Make 3.80 in
>>>> http://lists.xenproject.org/archives/html/xen-devel/2016-01/msg02134.ht 
>>>> ml 
>>> "decided" is a bit strong. I suggested these values. And while I'm
>>> pretty certain that even plain make 3.80 will work, I'm in no way
>>> sure plain 2.16.1 will (what I'm building with once in a while is some
>>> 2.16.9x, and I can't say how many backports it has). So the
>>> question really is - did you test that things build with these?
>> Why would he have done, you suggested 2.16.1 with no hint that you thought
>> it might not be a reasonable version to use.
>>
>> TBH having rejected Doug's original proposal I would have said it was up to
>> you to specify the actual precise versions you think should be used, rather
>> than making Doug guess and leading him down blind allies by making
>> apparently authoritative suggestions which you secretly aren't actually
>> sure about yourself.
> To be honest it didn't even occur to me that someone might
> propose such a patch without verifying things actually build
> (unless using more cautious wording). Also note that in the first
> reply to the v1 patch I did refer to 2.16.9x (which imo has made
> clear that that's the lowest one I ever tested with recently), i.e.
> I don't think I've actively mislead him.
>
>> Anyway we could go round and round like this forever. What's wrong with
>> starting with this as a baseline and bumping it if it turns out to be a
>> problem in practice?
> Well, we certainly could (which would be in line with my second
> reply to v1), just that I'm not sure how much value such a doc
> addition then has. At the very least it should then say "no
> lower than 2.16.1, something slightly newer may be needed" or
> some such.
>
>>> Also I'm not sure 2.16.1 is going to be sufficient for ARM (it's
>>> most definitely too old for ARM64).
>> I suppose there is an implicit max(version, first version supporting arch).
>> I don't think we can really go into the level of detail needed for per arch
>> toolchain requirements.
> I'm afraid quite frequently "first version supporting arch" isn't
> good enough. If we know otherwise for ARM64, that's certainly
> fine.
>
>> I certainly don't know which version of either gcc or binutils is needed to
>> build either ARM variant.
> Well, again - what's that documentation addition then good for?

Ping?

It doesn't matter exactly which version we choose as a minimum, but it
*is* very important that the version is written down.  Our README file
is the correct place for this information to live.

IMO, its also fine to say "For x86, GCC $X and Binutils $Y.  For ARM,
GCC $J and Binutils $K".

~Andrew