XUtil: const-correctnes for strchrnul

[BenBE]

This contains multiple parts:

1. Fix a few warnings related to const-correctness and strchrnul
2. Fix the buffer overflow handling in LinuxProcessTable_readCGroupFile by handling the snprintf retur values
3. Actually advance our write location allowing for multiple entries to be read

[Explorer09]

How does this &buffer[eol - buffer]; not trigger warnings at all?

Update: I have a better idea to the problem:

size_t eol_offset = (size_t)(eol - (const char*)buffer);
buffer[eol_offset] = '\0';

[BenBE]

Based on my reading of the ISO C99 standard:

• &p[i] returns a pointer to the ith element in p using the type of p
• b - a returns uintptr_t in units of the pointed-to types for any a and b that point inside the same object (array) in memory, if the pointed-to types are compatible, and not incomplete; irrespective of cv-qualifiers.

[Explorer09]

b - a should return a ptrdiff_t, but if they are in different types a compiler may label it as a warning (a strict aliasing violation, maybe). A compiler wasn't told that the pointer returned from strchrnul is derived from the first argument passed into it, and so it doesn't know that b - a would work unless it can examine the body of strchrnul function.

Note: I don't like the constness thing at all. Part of the PR #1915 that I was recently doing had been dealing with -Wcast-qual warnings from GCC that over-warns things. (Not strictly a false positive, but GCC wasn't taught that execv() is special that it has to make an exception and cast away the const qualifier.)

[natoscott]

LGTM, thanks @BenBE