Strict c99 compilator switches

[lelf]

Why do you need a Ťstrict c99ť? GCC is now yet fully supports C99 (for example, [*]-arrays), so there's no any point to talk about Ťstrictnessť.

And -pedantic-errors and -Werror are just useless here, I anyway can use GCC extensions via extension .

I suggest switching to gcc -std=gnu99 — that is, С99 + GCC extensions.

[noix]

Why do you need a Ťstrict c99ż? GCC is now yet fully supports C99 (for example, [*]-arrays), so there's no any point to talk about Ťstrictnessż.

And -pedantic-errors and -Werror are just useless here, I anyway can use GCC extensions via extension .

I suggest switching to gcc -std=gnu99 — that is, С99 + GCC extensions.

It's not supposed to be more powerful -- it supposed to be, well, more strict.
It allows for contests that demand using explicit casts where necessary etc., and generally writing a cleaner program -- not a program that just works, despite a ton of warnings. Some contest-setters used it for contests for students learning C language.

[uau]

It allows for contests that demand using explicit casts where necessary etc., and generally writing a cleaner program -- not a program that just works, despite a ton of warnings.

Some of the warnings generated by those switches occur in clean programs though. Especially treating "pointer targets differ in signedness" as an error is unreasonable IMO. Trying to "fix" that warning by casts or otherwise usually makes the program less clean and makes it more likely that real errors are masked (since the cast will now silence warnings about types that are really incompatible).

A real C99 mode with "-std=gnu99" and without the silly errors would be useful since GCC does not accept "for (int i=0; ..." syntax without the standard switch.

[noix]

It allows for contests that demand using explicit casts where necessary etc., and generally writing a cleaner program -- not a program that just works, despite a ton of warnings.

Some of the warnings generated by those switches occur in clean programs though. Especially treating "pointer targets differ in signedness" as an error is unreasonable IMO. Trying to "fix" that warning by casts or otherwise usually makes the program less clean and makes it more likely that real errors are masked (since the cast will now silence warnings about types that are really incompatible).

A real C99 mode with "-std=gnu99" and without the silly errors would be useful since GCC does not accept "for (int i=0; ..." syntax without the standard switch.

I'm not convinced SPOJ would gain anything from adding too many new "languages". It would probably just create confusion.
I think C++ (g++) is close enough to gnu99. Just remember about the few differences (void* casts, no default type in declarations, having to "#include" etc.; it's good style anyway, IMO).

As for usability of "C99 strict" language in SPOJ -- some contest-setters believe it really valuable. I also think so: I've seen solutions of some students who didn't even know how to get rid of some really not-silly warnings, but didn't bother as the program worked. It might be ok in some contests, but not on a C learning course.