From mboxrd@z Thu Jan  1 00:00:00 1970
From: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Date: Thu, 20 Sep 2018 08:45:22 +0200
Subject: [Buildroot] [PATCH 1/1] ipsec-tools: needs host-bison
In-Reply-To: <87o9ctp5if.fsf@dell.be.48ers.dk>
References: <20180915095906.11558-1-fontaine.fabrice@gmail.com>
 <87o9ctp5if.fsf@dell.be.48ers.dk>
Message-ID: <20180920084522.1dea57e8@windsurf>
List-Id: <buildroot.busybox.net>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: buildroot@busybox.net

Hello Peter,

On Wed, 19 Sep 2018 23:00:08 +0200, Peter Korsgaard wrote:

>  > diff --git a/package/ipsec-tools/ipsec-tools.mk b/package/ipsec-tools/ipsec-tools.mk
>  > index ddae9a75f3..8672272538 100644
>  > --- a/package/ipsec-tools/ipsec-tools.mk
>  > +++ b/package/ipsec-tools/ipsec-tools.mk
>  > @@ -9,7 +9,7 @@ IPSEC_TOOLS_SOURCE = ipsec-tools-$(IPSEC_TOOLS_VERSION).tar.bz2
>  >  IPSEC_TOOLS_SITE = http://sourceforge.net/projects/ipsec-tools/files/ipsec-tools/$(IPSEC_TOOLS_VERSION)
>  >  IPSEC_TOOLS_INSTALL_STAGING = YES
>  >  IPSEC_TOOLS_MAKE = $(MAKE1)
>  > -IPSEC_TOOLS_DEPENDENCIES = openssl flex host-flex
>  > +IPSEC_TOOLS_DEPENDENCIES = openssl flex host-flex host-bison  
> 
> Hmm, shouldn't this use BR2_BISON_HOST_DEPENDENCY /
> BR2_FLEX_HOST_DEPENDENCY instead?

No, that's not what we have agreed so far. Our idea was that
BR2_BISON_HOST_DEPENDENCY and BR2_FLEX_HOST_DEPENDENCY would be used
only for the kconfig stuff in the kernel/u-boot and al, but that we
would keep using our own host-flex/host-bison for the rest, and
especially for target packages.

There are two reasons to that:

 (1) We can be pretty confident that the flex/bison stuff in kconfig
     will have been tested/exercised against a wide variety of
     flex/bison versions, so using whatever flex/bison version
     available on the host system is good enough. However, for the rest
     of the packages that use flex/bison, we can't be so sure, so
     having our own flex/bison version allows us to be sure that things
     "will work".

 (2) For target packages, using the system-provided flex/bison version
     means that the generated code can be subtly different between
     flex/bison versions, which makes the build non-reproducible. Hence,
     it is important to have our own fixed version of flex/bison. For
     host packages, that is probably less of a concern, but point (1)
     remains valid.

Do we still agree on this position, or do you have
counter-arguments ? :-)

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com