illumos gate

This should result in effectively no change in behavior or text, and so should be testable via wsdiff

Patrick Mooney wrote in #note-2:

This should result in effectively no change in behavior or text, and so should be testable via wsdiff

IIRC sometimes const-ifying will allow callers to be less conservative about saving state. It looks like, however, your changed functions aren't calling OTHER changed functions, so wsdiff or diffs of dis(1) output should confirm no-real-change.

Tested with wsdiff(1), and the only differences in the kernel text (besides the normal ACPI noise) were the git SHA and the CTF for the changed functions:

platform/i86pc/kernel/amd64/unix
NOTE: ELF .data difference detected.

10398,10401c10398,10401
< 0706520  \0  \0  \0  \0  \0  \0  \0   m   a   s   t   e   r   -   0   -
<         00000000 6d000000 65747361 2d302d72
< 0706540   g   9   5   1   4   a   b   4   4   6   5  \0  \0  \0  \0  \0
<         31353967 34626134 00353634 00000000
---
> 0706520  \0  \0  \0  \0  \0  \0  \0   i   l   1   4   6   7   8   -   0
>         00000000 69000000 3634316c 302d3837
> 0706540   -   g   f   b   0   1   e   5   0   4   7   c  \0  \0  \0  \0
>         6266672d 35653130 63373430 00000000

platform/i86xpv/kernel/amd64/unix
NOTE: ELF .data difference detected.

8640,8643c8640,8643
< 1105220  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0   m   a   s   t   e
<         00000000 00000000 6d000000 65747361
< 1105240   r   -   0   -   g   9   5   1   4   a   b   4   4   6   5  \0
<         2d302d72 31353967 34626134 00353634
---
> 1105220  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0   i   l   1   4   6
>         00000000 00000000 69000000 3634316c
> 1105240   7   8   -   0   -   g   f   b   0   1   e   5   0   4   7   c
>         302d3837 6266672d 35653130 63373430

kernel/amd64/genunix
NOTE: ELF .SUNW_ctf difference detected.

42900a42901,42902
> typedef au_buff_t token_t;
>
42921,42922d42922
< typedef au_buff_t token_t;
<
46858,46862c46858,46862
< extern index_t bt_availbit(ulong_t *, size_t);
< extern void bt_copy(ulong_t *, ulong_t *, ulong_t);
< extern int bt_gethighbit(ulong_t *, int);
< extern int bt_getlowbit(ulong_t *, size_t, size_t);
< extern int bt_range(ulong_t *, size_t *, size_t *, size_t);
---
> extern index_t bt_availbit(const ulong_t *, size_t);
> extern void bt_copy(const ulong_t *, ulong_t *, ulong_t);
> extern int bt_gethighbit(const ulong_t *, int);
> extern int bt_getlowbit(const ulong_t *, size_t, size_t);
> extern int bt_range(const ulong_t *, size_t *, size_t *, size_t);

For a sanity smoke-test, I also booted up a BE running these bits and dtraced the bitmap functions to see that they were being called. The shipped header showed the const versions in place.