]> git.ipfire.org Git - thirdparty/binutils-gdb.git/commit
libctf: fix slices of slices and of enums
authorNick Alcock <nick.alcock@oracle.com>
Tue, 11 Feb 2025 14:32:40 +0000 (14:32 +0000)
committerNick Alcock <nick.alcock@oracle.com>
Fri, 28 Feb 2025 15:13:24 +0000 (15:13 +0000)
commit5a1d8eca5c331edab4e424c2034685433efa4bf5
tree4490fa6e4e7e044f04f7b6a8393e7592b85d01b0
parentf1b636cf3f9d9502444edbd602624237e609979e
libctf: fix slices of slices and of enums

Slices had a bunch of horrible usability problems.  In particular, while
towers of cv-quals are resolved away by functions that need to do it, towers
of cv-quals with slices in the middle are not resolved away by functions
like ctf_enum_value that can see through slices: resolving volatile -> slice
-> const -> enum will leave it with a 'const', which will error pointlessly,
annoying callers, who reasonably expect slices to be more invisible than
this.  (The user-callable ctf_type_resolve still does not resolve away
slices, because this is the only way users can see that the slices are there
at all.)

This is induced by a fix for another wart: ctf_add_enumerator does not
resolve anything away at all, so you can't even add enumerators to const or
volatile enums -- and more problematically, you can't add enumerators to
enums with an explicit encoding without resolving away the types by hand,
since ctf_add_enum_encoded works by returning a slice!  ctf_add_enumerator
now resolves away all of those, so any cvr-or-typedef-or-slice-qual
terminating in an enum can be added to, exactly as callers likely expect.

(New tests added.)

libctf/
* ctf-create.c (ctf_add_enumerator): Resolve away cvr-qualness.
* ctf-types.c (ctf_type_resolve_unsliced): Don't terminate at
the first slice.
* testsuite/libctf-writable/slice-of-slice.*: New test.
libctf/ctf-create.c
libctf/ctf-types.c
libctf/testsuite/libctf-writable/slice-of-slice.c [new file with mode: 0644]
libctf/testsuite/libctf-writable/slice-of-slice.lk [new file with mode: 0644]