Or, I guess, the people having problems learning to code are the people who happen to be on ancient compilers. Maybe it's not so surprising after all.
Many C programmers cannot make valid judgments as to the severity of a warning ('warning: bla bla may ...' => does it or doesn't it? I wouldn't know)
Also, those programmers typically would not know how to change their code to get rid of warnings. If they do, or if their compiler shoves a solution in their face (for example, a compiler might suggest to do "if((y=f(x)))") to get rid of an 'assignment inside if' warning), they do not see the advantage of the extra work.
The end result is code that produces hundreds of warnings. If a code change brings in a new one, people don't notice and/or don't care.
Some IDEs make things worse by allowing you to filter out warnings and messages from their output, so that you can focus on compiler errors. That is good for good programmers (they will fix errors, compile again, and then study every single warning), but bad for those who are happy to have code that compiles at all, as they will run into 'do not show me warnings' mode.
Really, if a programmer can't fix warnings (and that includes taking the time to learn why they are there), what are you paying them for? My policy on projects I work on is to enable all warnings, treat them as errors, then only disable warnings after trying to fix them, researching them, and finally discussing them, and usually then only on a case by case basis (only occasionally do I eschew certain warnings altogether (such as -Waggregate-return for g++), and I leave them commented out with the reason they are disbled).
(Edit: a quick tip for gcc, BTW: use -fdiagnostic-show-option for easier search of a warning, and the name of it in case you have to pragma it out).
If the default settings of gcc included throuwing out warnings for every time a function is declared without parameters and missing the void, lots of people would complain, and loud. It's a very common construct.
Interestingly, adding -std=c99 oddly removes the warning about the missing return statement in main().
I can't find a way to warn about the empty param list in the prototype, but it's possible that it's just not considered "bad enough style" to warrant one.
C++ is, of course, more strict. Try to compile that with g++, and you get several errors. The compiler needs to know how to mangle the symbol when it first encounters it, so you can't have "unspecified" parameter lists anywhere.