Re: [gentoo-dev] [PATCH 1/3] multiprocessing.eclass: search also GNUMAKEFLAGS for --load-average

[Sam James <sam@gentoo.org>]

Florian Schmaus <flow@gentoo.org> writes:

> [[PGP Signed Part:Undecided]]
> On 25/07/2023 10.32, Michał Górny wrote:
>> On Tue, 2023-07-25 at 08:26 +0200, Florian Schmaus wrote:
>>> On 25/07/2023 06.50, Michał Górny wrote:
>>>> On Mon, 2023-07-24 at 20:57 +0200, Florian Schmaus wrote:
>>>>> Since --load-average may not be found in other Make implementations
>>>>> besides GNU MAKE, it is potentially found in GNUMAKEFLAGS and not in
>>>>> MAKEOPTS.
>>>>>
>>>>> Signed-off-by: Florian Schmaus <flow@gentoo.org>
>>>>> ---
>>>>>    eclass/multiprocessing.eclass | 6 +++---
>>>>>    1 file changed, 3 insertions(+), 3 deletions(-)
>>>>>
>>>>> diff --git a/eclass/multiprocessing.eclass b/eclass/multiprocessing.eclass
>>>>> index e55be636a02c..6489ecbb44a6 100644
>>>>> --- a/eclass/multiprocessing.eclass
>>>>> +++ b/eclass/multiprocessing.eclass
>>>>> @@ -1,4 +1,4 @@
>>>>> -# Copyright 1999-2022 Gentoo Authors
>>>>> +# Copyright 1999-2023 Gentoo Authors
>>>>>    # Distributed under the terms of the GNU General Public License v2
>>>>>       # @ECLASS: multiprocessing.eclass
>>>>> @@ -86,7 +86,7 @@ makeopts_jobs() {
>>>>>    # @FUNCTION: makeopts_loadavg
>>>>>    # @USAGE: [${MAKEOPTS}] [${inf:-999}]
>>>>>    # @DESCRIPTION:
>>>>> -# Searches the arguments (defaults to ${MAKEOPTS}) and extracts the value set
>>>>> +# Searches the arguments (defaults to ${MAKEOPTS} ${GNUMAKEFLAGS}) and extracts the value set
>>>>>    # for load-average. For make and ninja based builds this will mean new jobs are
>>>>>    # not only limited by the jobs-value, but also by the current load - which might
>>>>>    # get excessive due to I/O and not just due to CPU load.
>>>>> @@ -95,7 +95,7 @@ makeopts_jobs() {
>>>>>    # If no limit is specified or --load-average is used without a number, ${inf}
>>>>>    # (defaults to 999) is returned.
>>>>>    makeopts_loadavg() {
>>>>> -	[[ $# -eq 0 ]] && set -- "${MAKEOPTS}"
>>>>> +	[[ $# -eq 0 ]] && set -- "${MAKEOPTS} ${GNUMAKEFLAGS}"
>>>>>    	# This assumes the first .* will be more greedy than the second .*
>>>>>    	# since POSIX doesn't specify a non-greedy match (i.e. ".*?").
>>>>>    	local lavg=$(echo " $* " | sed -r -n \
>>>>
>>>> I'm pretty sure [GNU]MAKEFLAGS has incompatible format, in particular it
>>>> makes hyphens optional.
>>>
>>> Yes, hyphens are optional in GNUMAKEFLAGS.
>>>
>>> However, makeopts_loadavg() would still be able to extract the
>>> hyphen-prefixed short (-l) and long (--load-average) options from
>>> GNUMAKEFLAGS. Hence having makeopts_loadavg() also inspect GNUMAKEFLAGS
>>> seems like an improvement over the current situation.
>>>
>> Also, shouldn't you handle MAKEFLAGS then as well?  If we're to
>> support
>> arbitrary variables used by build systems.
>
> We could.
>
> But GNUMAKEFLAGS was not arbitrary chosen. The idea is that portage
> may set --load-average via GNUMAKEFLAGS if the user did not set
> MAKEOPTS and GNUMAKEFLAGS.
>
> See https://github.com/gentoo/portage/pull/1072
>
> I first put --load-average into MAKEOPTS, but --load-average is not a
> portable make option, that is, some Make implementation do not support
> it. Adding it to GNUMAKEFLAGS, a variable already set by portage, we
> avoid passing this option to a make implementation that does not
> support it.
>
> Hence, just adding GNUMAKEFLAGS is sufficient for the purpose of
> propagating portage's potential new default into
> multiprocessing.eclass & Co.

But this exposed a problem in that we're missing other variables
that make recognises, so we need to handle the other case too.