#2 Improve "const correctness"

[Markus Elfring]

Would you like to apply the advices from the article
"http://en.wikipedia.org/wiki/Const_correctness" to
your sources?
I suggest to add the key word "const" to the type
specifiers for parameters like the following.
- msg (function "ipmi_request"), addr (function
"ipmi_validate_addr")

http://cvs.sourceforge.net/viewcvs.py/openipmi/OpenIPMI/include/linux/ipmi.h?rev=1.6&view=markup

- heap (function "print")

http://cvs.sourceforge.net/viewcvs.py/openipmi/OpenIPMI/unix/heap.h?rev=1.1&view=markup

[Corey Minyard]

Logged In: YES
user_id=236278

Ok, I've done this in the 2.0 releases. I'm not sure I'm
willing to backport this to 1.4, though, as the side-effects
may be too much.

[Markus Elfring]

Logged In: YES
user_id=572001

Are you going to check other files in the current release, too?

[Corey Minyard]

Logged In: YES
user_id=236278

I'm not sure what you mean. The changes are checked in to
the CVS head, but not into the 1.4 branch. There's a few
more things I want to deal with that are internal, and a
backwards compatibility problem with posix_vlog that I'm
going to fix a different way.

I'll look at doing just the changes you wanted to 1.4.

[Markus Elfring]

Logged In: YES
user_id=572001

I pointed just to two files for review.
Do you know more candidates for improvement already?

[Corey Minyard]

Logged In: YES
user_id=236278

Does the attached patch do what you suggest?

[Corey Minyard]

[Markus Elfring]

Logged In: YES
user_id=572001

Yes, this change list looks great. (Are the files "ipmi.h"
and "heap.h" excluded?)
I hope that new functions will be const-correctly designed
and developed from the beginning.

[Corey Minyard]

Logged In: YES
user_id=236278

Yes, I will keep this in mind for future updates. Thanks.
The 1.4.18 release has these changes.

ipmi.h is a kernel interface and heap.h is an internal
interface, so they are no as important. There's a few other
things, too, that need fixing. I'll do those over time,
probably.

[Markus Elfring]

Logged In: YES
user_id=572001

I suggest that this issue should be kept open until all
files will be fixed without exception.

[Markus Elfring]

Logged In: YES
user_id=572001

Do you notice any difference in the generated object code if
all proposed refactorings will be applied?

[Corey Minyard]

Logged In: YES
user_id=236278

I didn't look to see if anything changed. From this, I kind
of doubt it made much difference. The compiler can make
more optimizations with const types, but almost all the uses
of this are "Put the data into this container and pass it
in/on and don't use it any more". Not many new
optimizations that could be done there.

[Markus Elfring]

Logged In: YES
user_id=572001

Are you interested to know any measureable effects like
object file sizes if a const-correct design will become
available for your software?

[Corey Minyard]

Logged In: YES
user_id=236278

It would be interesting to know. If it made a difference,
it would be incentive to keep going :-).

[Markus Elfring]

Logged In: YES
user_id=572001

Would you like to join an other discussion?
http://groups.google.de/group/comp.lang.c/tree/browse_frm/thread/ed3e9f3bf137804e/171eed12e1b2c4c5

[Corey Minyard]

Logged In: YES
user_id=236278

I've gone over all the interface and I'm try to use const
whenever possible. Thanks for the heads-up.

[Markus Elfring]

Logged In: YES
user_id=572001

Thanks for this accepted correction.

Would you like to add a reference (case number) to this
issue to the change log or any comments?

[Corey Minyard]

This has been a while and I missed the last comment. I assume everything is ok, I'm not sure there's anything else I need to do.