Fix various GCC warnings #79
Conversation
|
There are a number of warnings in the Mac OS CI build yet. Some of these aren't as clean to fix nicely - I'm looking at the hidden virtual method warnings. The same warnings affect OpenBSD (which we have users on) and presumably other BSD variants. I am happy to fix these, but I do not want to burden the devs or go against the project goals. Please advise. Thank you! |
|
Clang issues those warnings also. |
|
I am not planning to add any more new fixes. This is everything that I can test on my Debian dev workstation. |
|
Can you split/squash this into two commits? One that only removes unused code and one with the rest. |
|
Is there a convenient way to do this in git or do I have to manually recommit the changes? I will force push once the branch tip is re-organized. |
|
If you have commits separated so that each would fit into either of the final commits you can use |
renderLine16_Blend(), renderLine16_Decal(), renderLine32_Blend() renderLine32_Decal() functions are now unused. remove clipLine() Another unused function. remove frand() The frand() computation implicitly cast the rMax integer to a float and in doing so unintentionally increased it. frand() is not used by us, so deleting it is a good solution. The entire randomizer could possibly be removed, but that is not within the scope of this commit. remove unused private fields I have removed unused private fields and their constructors, and one unused constant. remove unused software driver functions drawRectangle() and drawLine() have been removed, because IrrlichtMt's software drivers have been removed.
Overloading a method in a subclass hides the definitions of the superclass. I don't think this is causing an issue because the subclass explicitly overrides both of the superclass methods, but now there is no question. There are some vptr errors ubsan complains about which this may be related to, but I do not know. handle missing enumeration values in switch ESNRP_LIGHT and ESNRP_SHADOW were not handled in the scene node registration. They are now explicitly ignored. remove extraenous semicolons always have defaults in color converter switch The defaults were enabled only in release. Rationale for changing this is that the behavior is the same whether or not default is explicit. So disabling defaults for debugging is not really helpful, except that it displays a warning. fix root cause of stringop warning After researching the cause of the warning thoroughly, I realized that GCC is noticing the array[length] = 0 anti-pattern, where it expects array[length - 1] = 0. Sure enough, setting an intermediate variable to hold the allocation length (including the null terminator) lets us use the standard idom, which GCC has no qualms with.
JosiahWI
force-pushed
the
fix/warnings
branch
from
311c7b7
to
395ee7e
Compare
Minor code quality improvements. The main feature is that some unused functions were removed.