[Boost-testing] Re: V2 testing: static or shared

From: Vladimir Prus (ghost_at_[hidden])
Date: 2005-04-18 02:46:09

Alisdair Meredith wrote:

> [hoping I can finally mail the list]
> 
> Vladimir Prus wrote:
> 
>> This raises two questions about regression testing with V2:
>> 
>> 1. Should we use static linking by default, or shared linking. Shared
>> linking might decrease link times, and binary sizes, but probably not
>> all libraries are prepared for shared linking (especially on Windows)
> 
> I think this might need to be configured per-toolset.
> 
> For instance, there are many bugs in the Borland runtime libraries, so
> experienced Borland developers simply statically link all the time - it
> is not worth trying to find all the workarounds.

Does this mean static linking to *runtime*, or complete avoidance of shared
libraries? I'd find it surprising that borland does not allows to work with
DLLs.
 
> Likewise for debug/release builds.  There are some very nasty bugs in
> Borland exception handling that show up only in release mode
> (specifically, when inline function expansion is enabled)  There is no
> workaround for these, other than to throw the compiler switch.
> 
> I would hate to force these Borland settings on the other compilers,
> but currently there is not much to be gained testing Borland in other
> configurations.
> 
> Hence, testing should (if possible) be configured by toolset.

I think it's possible. Say, we can add

   <toolset>borland:<link-runtime>static

to Boost-wide requirements to force static linking to runtime on borland.

- Volodya