Re: [boost] [Removing support for old compilers]

Subject: Re: [boost] [Removing support for old compilers]
From: Edward Diener (eldiener_at_[hidden])
Date: 2014-01-13 10:35:58

On 01/13/2014 01:31 AM, Adder wrote:
>> Nevertheless, I still try to not break things under Borland, but that's kind
>> of hard because we don't even have a Borland tester. (And there isn't - to
>> my knowledge - a free version.)
>
> A very quick guide to using a free version of the Borland C++ Compiler
> (5.5) toolset may be found at the following address:
>
>    http://adder.demo.iworks.ro/Go/2014-01-12_Borland/
>
> It includes a few patches to the Standard Library implementation that
> are quite important (as demonstrated by the also-included mini test
> programs). As I have used them for a long time, I had almost forgotten
> about them.
>
> (Similar patches for std::auto_ptr are needed for all versions of
> Borland/CodeGear/Embarcadero including 2013 aka XE5, whethey they use
> RogueWave, STLport or Dinkumware. I also have versions for 2002 aka
> Builder 6, 2006 and 2007 and I am going to upload them too.)
>
> The mini-guide also describes a work-around for a long-standing bug in
> the linker.
>
> More recent versions can be used similarly, by downloading the trial
> versions from the CodeGear/Embarcadero website and then using the
> command line tools included. At least in older versions, the command
> line tools were not bound by the trial time limit.

And after the trial versions run out, how do you propose to test these 
compilers ?

>
> I am going to describe the modifications to borland.jam and other
> files that have proven helpful for me (e.g. to support DEF (module
> definition) files being passed to the linker) as soon as I have more
> time...

I worked with Borland C++ and C++ Builder (3,4,5,6) a number of years 
ago. Given Borland's near total inability to fix longstanding bugs over 
the years I will never work with Borland/Codegear tools again, nor will 
I ever spend a nanosecond of time attempting to make any library I 
create work with Borland/Codegear compilers. Microsoft may be bad in the 
offhand way in which they refuse to fix or address C++ bugs which they 
have decided are not mainstream enough to spend time on but 
Borland/Codegear was truly hopeless. Others can do whatever they want of 
course but this is my own personal experience.

I do not believe, as opposed to your own opinion, that implementors of 
Boost libraries should spend unnecessary time trying to accomodate 
ancient compilers. If the modern compilers are testable and have at 
least adequate support, then I can understand trying to work with them.