Re: [boost] Factoring out Test Library wrapstrstream

From: Fernando Cacciola (fernando_cacciola_at_[hidden])
Date: 2002-11-27 13:50:16

----- Original Message -----
From: "Rozental, Gennadiy" <gennadiy.rozental_at_[hidden]>
To: "'Boost mailing list'" <boost_at_[hidden]>
Sent: Wednesday, November 27, 2002 3:01 PM
Subject: RE: [boost] Factoring out Test Library wrapstrstream

> > The problem is that BCC, by default, uses an 'intrinsic'
> > version of strcmp
> > which is not a real function thus it is not located in std.
> > Declaring strcmp
> > inside std doesn't help.
> > The workaround is to either use unqualified calls to strcmp
> > or to disable
> > its intrinsic version. My patch does the latter.
>
> Why don't I have any problems compiling using borland command line tools?
>
Borland's intrinsic functions are enabled when the option 'Fastest Possible
Code optimization' is used.
This corresponds to the -O2 command line switch.
The alternative optimization, "Smallest Possible Code', which corresponds to
the -O1 switch, doesn't set intrinsic on.

My guess is that the jamtools for Borland uses -O1 instead of -O2.
The problem is that when you build the test by hand you don't know that you
must use -O1 instead of -O2.
I think it is better to use the pragma to turn it off than to require users
to use -O1.

Fernando Cacciola