[gdal-dev] Re: Low Level Library

[Stephane Villeneuve]

Hi,

I agree with the principle to have a stand alone library for many modules,
but keep in mind that all
users of this library should be programmer. 

I'am not very hot to add some ifdef in the source code if the memory stuff
is present or not, 
Why do not force the user of the library to link with a stubfile if it
don't want to use the real library.

I think,the best approche is to be easy for the user to use is stuff of
memory without make modifications
in many files. 

The switch case should be in the Makefile stuff( with some configuration
file)

link with the memory stub or with the real memory management, and the user
could modify the stub
file to use is memory system. The same think could be done for other
module.

The release package could include the both. 

The makefile could access a file to know which file is include of the
complete memory system and a other to know
which files are included in the stub memory stuff. They could share the
smae directory. and the makefile could use
one.

memory.dir
stubmemory.dir



About Error handle, I think that it should be interresting if we don't only
handle the last error, but all errors, A fatal 
error should be occur because you receive a warning error 3 call before,
And the user or the programmer should be 
interested to see all errors. I don't really like a printf everytime a
warning occur. it shoul be interesting that the 
caller functions or the GUI take many errors to create only one.

Is a Fatal error equal a programmer error??

> > I never used CVS before... does it allow remote checkin/checkout? I
> > think I already read something that made me believe it did...

I'am not really sure that it's a good idea to provide remote checkin to
anyone. How do you think to drive 
the checkin/checkout security???

I'am not remember if it's for linux or a other GNU package, but the checkin
system is :

you download the developpement library you make some work,you send your
modifications to the source master in a RCS-DIFF format with the devlib
version you use, they will check your modification and they will include
them.

Error class, 

I think that all error number and error string should be created in the
same file, to provide some NLS lookup stuff.

CONV_ERR_BASE 1200
GDAL_ERR_BASE 2000

GDAL_ERR_BASE+1 = "Unknown Format"

this complicate the error manipulation but it create a error text standard.

> === message truncated ===
I don't know what append, I lost a part of the Danmo mail. :-(


regards,
stephane
------------------------------------------------------------------------
Free Web-based e-mail groups -- http://www.eGroups.com