Fix an incorrectly aligned result when converting between pointer types

[derselbst]

Clang-3.x raises the following warning:

https://travis-ci.org/FluidSynth/fluidsynth/jobs/450156015#L991

This is because the char idlist[] is accessed like a 32bit integer, but as being a character array, it may not be aligned as such. This results in undefined behaviour. To fix it I put it in a union along with a uint32 telling the compiler to suitably align it for integer access. Now, the warning is gone:

https://travis-ci.org/FluidSynth/fluidsynth/jobs/448425040#L841

@mawe42 Do you agree with this fix?

[mawe42]

That looks good! Nifty trick to use a union with a 4-byte integer to force proper alignment. But maybe it would be good to add a short comment above the union to explain it's purpose?

[jjceresa]

Cannot we use "const char c[ ]" instead of "const char c[116]" ?

[derselbst]

Sure, comment added.

Cannot we use "const char c[ ]" instead of "const char c[116]" ?

No. In C89, arrays wraped in unions must have a fixed size. Otherwise the size of the union would depend on the initialization of its first member, which results in different sizes for different instances of the same union type.

[mawe42]

Very good comment, thanks. Two small typos noted.

[mawe42]

accessed

[mawe42]

comparison

[jjceresa]

Otherwise the size of the union would depend on the initialization of its first member, which results in different sizes for different instances of the same union type.

Oh, thanks , I had forgotten for this !

Next is out of subject: Also, sorry for the absence of "permanent link" , "quote" and "formatting" message. Since today I cannot use these possibilities anymore (these are now disabled). Guys at GitHubSoldOffByMicrosoft now have complicated things. Here GitHub web page was notified requesting for upgrading another recent web browser or OS !.

[derselbst]

No problem jjc :)