Re: [lmi] Fixing build with clang 7 and 64-bit Linux

[Vadim Zeitlin]

On Sun, 11 Nov 2018 15:17:32 +0000 Greg Chicares <address@hidden> wrote:

GC> On 2018-11-02 16:39, Vadim Zeitlin wrote:
[...]
GC> >  The first two errors, about the unused private fields, actually do also
GC> > look straightforward to me and I'd just remove these member variables as
GC> > well as the corresponding SequenceParser ctor arguments. But, AFAIR, the
GC> > last time this warning came up, you preferred to suppress it rather than
GC> > remove the unused fields.
GC> 
GC> Yes. It came up for exactly these two data members, and I worked around
GC> the clang diagnostic; but later I removed the workaround, thinking that
GC> it was intended for some obsolete compiler. I've re-added the workaround.

 Oops, sorry, I completely forgot that this had already happened with
exactly those members, otherwise I would have been able to track down the
changes myself, but I didn't even try to do it because I thought these
warnings were new.

GC> Instead, I've embraced 'int' as the one and only universal integer
GC> type for the twenty-first century; lmi doesn't need arrays so big
GC> that 'int' can't index them.

 FWIW I agree with this, if a larger type is really needed, we ought to use
int64_t or int128_t explicitly.

GC> I've done it, and will push soon.

 Thank you, I can confirm that with these changes and a few more hopefully
non-controversial ones, as they mostly just add preprocessor checks for
clang to avoid using gcc pragmas that it doesn't know about it with it, lmi
builds and runs when using this compiler. These "few more changes" are at

        https://github.com/vadz/lmi/pull/103

Please let me know if you see anything wrong with them.

GC> CI with other compilers is a useful complement to my habit of running
GC> all tests before any push.

 I'll try to (finally) set it up for lmi soon, at least under Linux. Ideally
I'd like to do it for MSW too but things are more complicated there...

 Thanks again for the fixes,
VZ