[tlbuild] -funsigned-char is not portable

[Jonathan Kew]

On 3 Jun 2009, at 15:02, Peter Breitenlohner wrote:

> On Wed, 3 Jun 2009, Jonathan Kew wrote:
>
>> Surely the correct fix is to ensure that the code does not depend  
>> on whether chars are signed or unsigned by default, as IIRC the C  
>> standard says this is implementation-defined, and there's no  
>> guarantee that every compiler even has an option to control it. If  
>> the code requires that char-sized variables are unsigned, it should  
>> simply declare them as "unsigned char", not assume it can rely on a  
>> compiler option.
>
> Hi Jonathan,
>
> you are absolutely right. But are you willing to check the bibtex8  
> code and
> adapt it if necessary at this late stage of building for TL2009?

No, I'm afraid not... and I don't expect that you will either. But  
perhaps it'd be worth reporting the issue to whoever is the bibtex8  
maintainer. (Hmmm, is there one?)

> I'd love to have a small bibtex8 test case for 'make check', that  
> would also
> detect if the programs really depends on char being unsigned.
>
> As I just started to write to Vladimir, this flag was present in the  
> TL2008
> version of Makefile.in for bibtex8.  When I translated that into  
> Makefile.am
> for TL2009 I first overlooked/forgot this flag and only recently
> reintroduced it.
>
> In other words, this flag was used in the TL2008 build and I dislike  
> to just
> drop it without further analysis.
>
> So how to proceed?

I suppose we keep the flag for now, with whatever configure-time magic  
is needed to handle non-GNU compilers, but it's something that would  
be good to note and get fixed in the future.

JK