I don't understand why radix was used in DIBuilder. One byte is always 8 bits, and LLVM already assumes that, so there is no point in making it variable.
And there definitely isn't 64 bits in a byte (the width of an intptr). This means the width of types was 8 times larger than they should be: an int32_t was considered 256 bits instead of 32 bits (cf the unit tests, where its size was also erroneously set to 8 bytes).
I don't understand why
radixwas used in DIBuilder. One byte is always 8 bits, and LLVM already assumes that, so there is no point in making it variable. And there definitely isn't 64 bits in a byte (the width of an intptr). This means the width of types was 8 times larger than they should be: an int32_t was considered 256 bits instead of 32 bits (cf the unit tests, where its size was also erroneously set to 8 bytes).Fixes #239