Re: [Tinycc-devel] modern c++ compiler written in C

[Christian Jullien]

I understand the need for LTS but, unfortunately C++ never worked like this.
I have been very happy with CFRONT 2.0, then with 3.0 when it came out but
I'm no more able to use them for my developments.

I recognize that some projects can be very happy to use only features from
CFRONT 2.0 only.

Except for 'small' projects you generally need to compose your application
with other C++ library and here comes the nightmare. 
Latest C++ is generally like a 'virus'. Because, for good or bad reasons,
you absolutely need the latest C++ library you depend on, you also need to
move your code to the same C++ version.
At the opposite, a C++ library using only CFRONT 2.0 features will not be
usable by more recent C++ versions - I mean C++ < 10 years!!

It is especially true when you develop on different platforms.

LTS C++ will remain a dream.

Again, except for the fun (life is not interesting without fun), I see no
interest for cfront. I bet that if we have the rights to use cfront (say
3.0) you soon will have a zillion of requests to add 'this important C++
feature I can't leave without'.

I'll stop to pollute this mailing list with this (interesting) thread.

Christian 

-----Original Message-----
From: address@hidden
[mailto:address@hidden On Behalf Of
address@hidden
Sent: vendredi 15 mai 2015 10:03
To: address@hidden
Subject: Re: [Tinycc-devel] modern c++ compiler written in C

On Fri, May 15, 2015 at 08:50:44AM +0200, Christian Jullien wrote:
> Of course, I know C++ programs that satisfy themselves with 'basic' 
> C++ features. The difficulty is to agree on 'basic' features. No two 
> C++ users use the same 'basic' features. Some want only encapsulation +
ctor and dtor.
> No more. Others want to add inheritance. 

Basically this means that different people and tasks need different
languages. The intention of "one size is to fit all / to cover all needs"
leads to a very complex solution.

Given that there are needs and people satisfied with C++ without templates
and exceptions, it would be nice to keep that "pre-"standard and
implementation alive.

I have one extremely important application, the Coda file system, which
belongs to this class. The complexity of modern C++ compilers is several
orders of magnitude bigger than of the application itself.
This is bad, for several reasons (begin with security...).

Another very useful tool, FLTK, looks to be compatible with cfront and also
is much less complex than the modern development tools.

So it is not wrong per se to have a complex language, but the fault is to
"pretend" that the next C++ version becomes better for all purposes and is
expected to be the only available/used one.

Strictly speaking each version is a definition of a new language, but an
appearance of a new language does not render the other ones obsolete. The
versions are treated like they do obsolete the previous ones, which is
actually not true.

Sergey Korshunoff wrote:
> Solution? May be a small c++ compiler (cfront-1 .. cfront-3) which can 
> be considered LTS version.

Sounds just right.

Rune


_______________________________________________
Tinycc-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/tinycc-devel