Re: [PATCH v2 1/2] wrapper: add a helper to generate numbers from a CSPRNG

From: Junio C Hamano <gitster@pobox.com>
Date: Jan 4, 2022, 21:39 UTC
Message-ID: <xmqqlezvb2p2.fsf@gitster.g>
In-Reply-To: <xmqqsfu3b4gw.fsf@gitster.g>

Junio C Hamano <gitster@pobox.com> writes:

Show 28 quoted lines

>> +ifeq ($(strip $(CSPRNG_METHOD)),arc4random)
>> +	BASIC_CFLAGS += -DHAVE_ARC4RANDOM
>> +endif
>> +
>> +ifeq ($(strip $(CSPRNG_METHOD)),arc4random-libbsd)
>> +	BASIC_CFLAGS += -DHAVE_ARC4RANDOM_LIBBSD
>> +	EXTLIBS += -lbsd
>> +endif
>> +
>> +ifeq ($(strip $(CSPRNG_METHOD)),getrandom)
>> +	BASIC_CFLAGS += -DHAVE_GETRANDOM
>> +endif
>> +
>> +ifeq ($(strip $(CSPRNG_METHOD)),getentropy)
>> +	BASIC_CFLAGS += -DHAVE_GETENTROPY
>> +endif
>> +
>> +ifeq ($(strip $(CSPRNG_METHOD)),rtlgenrandom)
>> +	BASIC_CFLAGS += -DHAVE_RTLGENRANDOM
>> +endif
>> +
>> +ifeq ($(strip $(CSPRNG_METHOD)),openssl)
>> +	BASIC_CFLAGS += -DHAVE_OPENSSL_CSPRNG
>> +endif
>
> Use of $(strip ($VAR)) looks a bit different from what everybody
> else does with ifeq in this Makefile.  Was there a particular reason
> to use it that I am missing?

Another thought. As far as I can see on the C code side in the later part of the patch, we are prepared to see multiple HAVE_* for CSPRNG defined by the builders, and let us choose the best one for them. I wonder if it makes sense to allow

    make CSPRNG_METHOD='arc4random getentropy'

as a way to tell us that they have these two and want us to pick the best one for them.

It does not add much value for human builders, but I suspect that it would make it simpler when we need to add autoconf support.

If we do not allow multiple methods listed on the CSPRNG_METHOD variable, the configure script will be forced to pick one in some way when multiple methods are possible on the platform, either by detecting all and picking one, or detecting serially from most preferred and stopping at the first hit.