I have hit the issue
before where object files or libraries passed to the linker had to be in a
certain order, it was because the linker was a single pass (as opposed to two
pass) linker which had to be able to resolve all symbols in order without
scanning through everything first. I don't believe this is that common
anymore, but this may be something the build mechanism being worked on may
need to support.
Cory
Bialowas
Rational
Software Corporation
Product
Manager
Phone: +1
(613) 591-2906
Mobile: +1 (613)
852-2073
Email: cbialowas@xxxxxxxxxxxx
-----Original
Message-----
From: Schaefer,
Doug [mailto:dschaefer@xxxxxxxxxxxx]
Sent: Tuesday, December 17, 2002 9:46
PM
To:
'cdt-dev@xxxxxxxxxxx'
Subject: RE: [cdt-dev] New build
paradigms
Actually, object files can be created in any order
since the references aren't resolved until link time. But that does
bring up a great point about the order of the files on the link line since
some of the linkers fail if you didn't have references before definitions (or
visa versa, I can't remember the exact scenario, I've been spoiled too much
lately by Visual Studio's automated build ;). But, that's a whole other
area that will need to be investigated for automated
build.
Doug Schaefer
Senior Staff Software Engineer
Rational - the software development
company
Ottawa
(Kanata), Ontario, Canada
-----Original Message-----
From: Alain Magloire [mailto:alain@xxxxxxx]
Sent: Tuesday, December 17, 2002 7:34
PM
To:
cdt-dev@xxxxxxxxxxx
Subject: Re: [cdt-dev] New build
paradigms
>
> This message is in MIME format. Since your mail
reader does not understand
> this format, some or all of this message may not
be legible.
>
>
------_=_NextPart_001_01C2A533.4217B400
> Content-Type: text/plain
>
> I actually don't think (1) is that hard to
do. Yes, you do need to figure
> out the dependencies between translation units
and the header files they
> bring in. This information would also be
necessary to generate proper
> makefiles that do incremental build.
However, it shouldn't be too difficult
> to do. Gcc provides an option, -M, to
generate this for you and I've build
> a perl script in the past to does the same.
But, as Chris mentioned, the
> CDOM, or more properly, the indexer, should be
able record what header files
> a particular file includes for a particular set
of build settings. But that
> will take some time to put in
place.
Yes, and probably it will be possible to go further in
finding the
dependencies. Most C programmers(sic) do not
even bother to put
the includes are to declare the functions
before:
# cat main.c
int main() {
int x = 0;
printf("Hello world");
/* define in file foo.c */
foo();
return 0;
}
The C parser by looking at the file can probably
figure out
that
there is a dependency on foo.c and tell the builder that it
needs
to compile
foo.c before main.c. Usually this is done manually in
a Makefile
main : main.o foo.o
$(CC) -o main main.o foo.o
main.o: foo.o
_______________________________________________
cdt-dev mailing
list
cdt-dev@xxxxxxxxxxx
http://dev.eclipse.org/mailman/listinfo/cdt-dev
