[Buildroot] [git commit] toolchain-external: update Linaro ARM toolchain

Thomas De Schampheleire patrickdepinguin at gmail.com
Fri Oct 11 09:16:33 UTC 2013


Hi,

On Fri, Oct 11, 2013 at 11:01 AM, Luca Ceresoli <luca at lucaceresoli.net> wrote:
> Thomas, Arnout,
>
>
> Thomas Petazzoni wrote:
>>
>> Dear Arnout Vandecappelle,
>>
>> On Thu, 10 Oct 2013 08:53:49 +0200, Arnout Vandecappelle wrote:
>>
>>>    I believe I asked this before: do we really want to remove external
>>> toolchains without going through a deprecation cycle? Why do we offer in
>>> a released buildroot three different Linaro toolchain versions that are
>>> all gcc 4.8 and that you cannot use anymore three months later?
>>
>>
>> Linaro toolchains are released every month, so they are moving quickly.
>> What are you proposing to do to handle this?
>>
>> Have just a BR2_TOOLCHAIN_EXTERNAL_LINARO_ARM option, which maps
>> automatically to the latest version of the toolchain, like we do for
>> packages?
>>
>> Or do you suggest to preserve toolchains? For how long?
>>
>> Until now, the arbitrarily chosen policy was to keep only three
>> versions of a given toolchain. For Sourcery CodeBench toolchains, I
>> believe it works quite well because the frequency of releases is not
>> too high. However, it's true that for Linaro toolchains, it may not be
>> appropriate, but I'm not sure what to do exactly.
>
>
> I also would like toolchains to be maintained in BR a little longer, at
> least one year.
>
> Of course maintaining 12 Linaro toolchains would be quite annoying and
> probably useless.
>
> So we may want to add only those Linaro toolchains that have relevant
> changes, say they upgrade gcc or provide some shiny new feature. But this
> may be risky: a newly introduced feature is more likely to be buggy. So...
> how about the 2nd or 3rd release after a relevant change? Uhm, quite fuzzy
> indeed.

The principle of keep three toolchains that differ sufficiently seems
ok to me. I don't think we should care about 'buggy' releases: Linaro
is responsible of delivering quality toolchains. If it happens that a
newly released toolchain is no good for some reason, we can still step
that one back, or take the newer one if it was released in the mean
time.

So, this proposal would for example have (fictional)
Linaro 2012.11 (based on gcc 4.4)
Linaro 2013.02 (gcc 4.4 but new glibc)
Linaro 2013.08 (gcc 4.8)

Best regards,
Thomas



More information about the buildroot mailing list