30 Jan
2021
30 Jan
'21
4:37 p.m.
It was thus said that the Great Bill Gunshannon via cctalk once stated:
On 1/29/21 4:12 PM, Will Cooke via cctalk wrote:
And the problem here is really quite plain and simple.
Why are you shipping code with any warnings?
Because sometimes they aren't. Example:
gcc -std=c99 -g -Wall -Wextra -pedantic -fPIC -g -shared -o lib/tcc.so src/tcc.c -ltcc
src/tcc.c: In function cclua_get_symbol':
src/tcc.c:528: warning: ISO C forbids assignment between function pointer and `void
*'
ISO C may forbid that, but POSIX requires it, and I'm compiling on a POSIX
system, but there isn't (to my knowledge) a way to state that. Yes, I could
probably surpresss that one warning, but for me, it's easier to ignore on
POSIX systems.
Also, have you tried clang with the highest warning level? It's useless.
I tried it once, only to be told "warning: struct foo has padding bytes
added" (or something to that effect). Okay, so I pack the structure, only
to get "Warning: struct foo doesn't have any padding". Yean, real useful
that.
-spc
On 01/29/2021 2:58 PM Fred Cisin via cctalk
<cctalk at classiccmp.org> wrote:
'=' and '==' makes possible what
is probably the most common error, and
which the compiler doesn't catch:
if (x = 3) . . . /* sets x to 3 and gives TRUE for the condition */
I imagine that there are probably some pre-processors that would return a
WARNING for it.
Modern Visual Studio and GCC both flag the "=" in a condition, I believe.
But if you're shipping code with 260+ warnings, who would see one more.