Two LSS talks [LWN.net]

Two LSS talks

Posted Oct 10, 2013 12:49 UTC (Thu) by etienne (guest, #25256)
Parent article: Two LSS talks

Another anti-patterns which seems to have changed behaviour with newer GCC, at least on ARM:

void fct (const char *str)
{
char *ptr = strchr (str, ':');
if (ptr != 0) *ptr = '\0';
... do something with str which is declared const ...
}

to post comments

Two LSS talks

Posted Oct 10, 2013 16:13 UTC (Thu) by mathstuf (subscriber, #69389) [Link] (6 responses)

Well, I don't see how this was guaranteed to work anyways. If `str` points to ro memory, you're hosed anyways. The C standard should fix these const -> non const issues at some point (as C++11/C++14 is finally doing). Is there a `strlchr` function floating around somewhere?

Two LSS talks

Posted Oct 10, 2013 19:39 UTC (Thu) by BenHutchings (subscriber, #37955) [Link] (5 responses)

The C standard should fix these const -> non const issues at some point (as C++11/C++14 is finally doing).

What you mean by that? Even C++98 requires std::strchr() and similar functions to be overloaded so that the return type matches the pointer argument type. I don't know how that could be achieved in C.

Two LSS talks

Posted Oct 10, 2013 20:06 UTC (Thu) by mathstuf (subscriber, #69389) [Link] (4 responses)

Well, the ideal fix's first step is "acquire time machine", but other than that, there's not much to be done other than something like a flag day (which I don't see happening any time soon), or adding overloading to C (somehow…), or CDR (C Done "Right") (which would never take off due to the incompatibilities).

If you're replying to the C++ side of things, it's most in reference to the fact that lots of container methods are now accepting const_iterator instead of iterator (such as erase() and insert()) so that you don't have to "infect" non-const-ness around to other places (or const_cast).

Two LSS talks

Posted Oct 11, 2013 9:24 UTC (Fri) by dgm (subscriber, #49227) [Link] (3 responses)

CDR exists. It's called Embedded C++: http://en.wikipedia.org/wiki/Embedded_C%2B%2B.

I always though ABC (A Better C) was a better name but, apparently, that's already taken.

Two LSS talks

Posted Oct 11, 2013 10:00 UTC (Fri) by khim (subscriber, #9252) [Link]

Embedded C++ is very strange creature because it removes one of the most important advantages of C++: namespaces. Thing like a namespace is more-or-less required to manage large programs and Embedded C++ forced the same fourty years old substitute (prefixes and/or suffixes).

Frankly I don't see a point. It's not really compatible with C++, but it's not compatible with C either. It includes not usable rounded-up set features but bunch of easily implementable features instead. In short: it's something designed to brag about, not something designed with real work in mind. Very similar to LSB: both in idea and in implementation. Only LSB has slowly started to acquire useful features recently, while EC++ just died.

Two LSS talks

Posted Oct 11, 2013 10:57 UTC (Fri) by HelloWorld (guest, #56129) [Link]

Embedded C++ is a pathetic idiocy that remove tons of useful functionality without any reason and doesn't solve any of the real problems that C and C++ have. Quoting Stroustrup: "To the best of my knowledge EC++ is dead (2004), and if it isn't it ought to be." (http://www.stroustrup.com/bs_faq.html#EC++).

Two LSS talks

Posted Oct 15, 2013 12:05 UTC (Tue) by jwakely (guest, #60262) [Link]

There is nothing done right about EC++, you'll have to try harder than that.