FreeTDS Development Group

[Frediano Ziglio <freddyz77 AT tin.it>]

Il ven, 2003-04-11 alle 05:17, James K. Lowden ha scritto:
> Freddy, 
> 
> configure.in: disable debug by default
> 
> Is this about assert()?  

Yes

> 
> When you change configure.in, you change the nightly snapshot and the
> distributions.  Maybe those would be good changes, but I think we should
> discuss them.  
> 
> I don't think --enable-extra-checks should mean "#define DEBUG 1".  I
> frankly don't know what it should mean.  I can see having a --debug and
> --release, or --debug={yes|no}.  
> 

You are right. Adding a --debug option is good. I though
--enable-extra-checks should enable expensive check on library (i.e.
check all internal structure for every call, or other thinks).

> I think as a developer's library in a state of high flux, it's good to
> have DEBUG defined in the nightly snapshots.  I think everyone -- users
> and developers alike -- is *much* better off that way.  Those interested
> in extreme performance can turn it off, when they're sure it will help. 
> At least they'll know they did so.  
> 

So, you suggest to add --debug option and enable it by default...

> I don't think assert() is expensive, certainly not the 4X problem you
> suggested.  I tried to demonstrate as much to you.  But I will refrain
> from using it in the ODBC library, in deference to you.  
> 

No, ODBC shouldn't be considered "the last wagon wheel" (as we use to
say in Italy)

> I don't claim to know all the answers.  I'd like to solicit other opinions
> and discuss the issue openly, in hopes of developing a consensus.  
> 
> Regards, 
> 
> --jkl

bye 
  freddy