Fixed assorted compiler warnings

[tdulcet]

• Fixed 20 stringop-truncation warnings.
• Removed use of the snprintf_nowarn macro, which revealed some hidden warnings.
  • Fixed 20 format-truncation warnings by doubling the size of the cbuf array.
• Fixed 10 discarded-qualifiers warnings.
• Updated the ASSERT function to (partially) use a macro instead, similar to the C standard library assert macro in assert.h.
  • This should reduce the number of false positive warnings, as compilers and static analyzers can much more easily check if the assert is going to fail.
  • I switched to Ernst's C99 ASSERT function, which shows the function name in error messages and I renamed it to ABORT. I also added the assertion string to the error messages.
  • This allowed removing the HERE argument, which I added to the macro instead. This was a major change, as Ernst had over 4,000 assert statements.
  • This should result in a performance improvement, as it no longer needs to call a function to test the expression.
• Updated to use the int8_t, int16_t, int32_t and int64_t integer types from stdint.h instead of redefining them.
  • This could help with the Windows issues. Unfortunately, to prevent more compiler warnings it required also updating over 700 of the printf calls to use the defines from inttypes.h.
  • I also used the boolean types from stdbool.h while I was at it.

This pull request is a work in progress, as I am not sure it is the best way to fix some of these warnings. I would recommend reviewing each commit individually. Feedback is welcome!

[xanthe-cat]

Just a general comment on one particular category you’ve looked at in this PR. Adding the tested condition for the ABORT message should be highly useful; Ernst may have had some idea what was going on but reading his code and the output it generates is often rather obscure to me. As much explanatory information as available is sometimes barely enough.

[tdulcet]

Adding the tested condition for the ABORT message should be highly useful

Thanks, that is good to know it would be useful for you. It is a feature of the standard library assert statement from assert.h, which I thought might be useful to add to Ernst's custom implementation. Unfortunately, until someone is able to find a fix for #10, I think I am going to have to rework this PR and remove some of the changes.

[ldesnogu]

Random comments:

• The last commit of that branch (adding const) introduced new errors when doing a non multi-threaded build.
• I think we should have an ASSERT_ALWAYS (or some other name such as FATAL); this would allow us to use the gcc/clang noreturn attribute.

Regarding the first comment above, I'm unsure how to proceed: amend or add a patch. My preference is to have bisectable code, which would mean amending the offending git change, instead of adding a new commit after.

For the second comment, this can come as a later patch.

[tdulcet]

Regarding the first comment above, I'm unsure how to proceed: amend or add a patch.

Please feel free to amend my commit. I changed the base of your PR, so it should automatically update when this is fixed (might need a quick rebase).

For the second comment, this can come as a later patch.

I previously moved the assert changes to #24, which does add the noreturn attribute, but this is currently blocked on fixing #10. Feel free to make additional changes in the PR if you want.

[ldesnogu]

Regarding the first comment above, I'm unsure how to proceed: amend or add a patch.

Please feel free to amend my commit. I changed the base of your PR, so it should automatically update when this is fixed (might need a quick rebase).

I force pushed the change. I will use this as a basis to fix other errors for single-thread compilation (if that proves possible at all, which I'm not sure is the case).

[ldesnogu]

-- ignore --

[ldesnogu]

This looks good to me.

The only thing I'd add is the removal of the snprintf_warn macro. This can be taken care of later and should not block the merging.