Remove unused free_codepages_list

[metux]

The free_codepages_list() function is only called on termination, where it's unnecessary, since the process will terminate a few cycles later.

(in general, those things aren't necessary on multiprocessing-platforms at all)

NAK'ed-by: slyfox

[ossi]

note that it makes sense to have cleanup functions in a conditional branch to simplify working with memory leak checkers.

[metux]

Replying to ossi:

note that it makes sense to have cleanup functions in a conditional branch to simplify working with memory leak checkers.

hmm, very good point, i hadnt considered.
perhaps we should add some --enable-memory-cleanup option which
causes this code to be enabled (otherwise it's just empty macros or inlines).

a perhaps more sophisticated approach might be using an garbage collection.

[andrew_b]

Replying to metux:

hmm, very good point, i hadnt considered.
perhaps we should add some --enable-memory-cleanup option which
causes this code to be enabled (otherwise it's just empty macros or inlines).

a perhaps more sophisticated approach might be using an garbage collection.

Absolutely no reason to introduce some strange options. All memory allocated in program must be freed in that program. Just use this simple rule and don't confuse other people.

[ossi]

Replying to metux:

perhaps we should add some --enable-memory-cleanup option which
causes this code to be enabled (otherwise it's just empty macros or inlines).

typically this is a run-time option, enabled via an environment variable (often, glibc's MALLOC_CHECK_ is "borrowed"). also, valgrind-dev provides a header file with a function to check whether the program is run under valgrind control.

a perhaps more sophisticated approach might be using an garbage collection.

bit of overkill for mc, huh? ;)

Replying to andrew_b:

All memory allocated in program must be freed in that program.

which the OS will happily do for us without even being asked ...
programs which do a lot of allocations (probably not mc) will be considerably slowed down on exit if they free everything explicitly.
on top of that, some allocators will cause basically all of the virtual memory to be swapped back into ram just for being able to free it, which will slow down the exit even more.

Just use this simple rule and don't confuse other people.

yes, life is so much simpler if one defines arbitrary rules and ignores the implications. ;)

[slyfox]

Correctness matters.

• mc is not the program, expected to be ran 10K times in a minute
• it will take additional cryptic unneeded code for leak checkers. "code for tool" is not nice. I like "tool for code" more :]
• If some handicap system wants it - devs can "patch" code themselves (they anyways will do that heavily)