[Swan-dev] true or TRUE?

[Andrew Cagney]

On 22 April 2018 at 14:07, Paul Wouters <paul at nohats.ca> wrote:
> I prefer TRUE over true

The linux kernel started switching to bool-true-false way back in '06
(2.6.x kernel?, >=RHEL5?) with:

commit 6e21828743247270d09a86756a0c11702500dbfb
Author: Richard Knutsson <ricknu-0 at student.ltu.se>
Date:   Sat Sep 30 23:27:11 2006 -0700

so, arguably, libreswan's klips code should be updated to follow the
not-so-current linux kernel coding style.  This would let us eliminate
the duplicate #define TRUE/FALSE in libreswan.h (leaving just the one
in constants.h).

Of course this also makes for a slippery slope.

>> On Apr 22, 2018, at 12:18, D. Hugh Redelmeier <hugh at mimosa.com> wrote:
>>
>>
>> Our current header system is a mess of several conventions and
>> non-conventions mingled.  That's something I would like fixed.
>
> And take into account KLIPS is slowly being phased out for the new xfrmi2 interfaces.
>
> Maybe split the shared KLIPS stuff out of libreswan.h ? But then again, I would wait for KLIPS to go completely before doing this cleanup.
>
>> If libreswan.h is not the correct baseline header, we should change
>> that #include appropriately.

"constants.h", which is really the concatenation of all the
"*constants.h" files, is slightly better.  However, it too has
accumulated all sorts of declarations and definitions that that are
not constants.  For instance, lset_t, enum_names, ...

Andrew