`NOTREACHED()` has been marked as `[[noreturn]]`, and therefore will
break in many cases when building the code. The migration variant of the
macro is being used upstream to allow for a gradual migration of each
callsite.
Chromium change:
https://chromium.googlesource.com/chromium/src/+/61831cb39155842c781f63214a7862bfd23e3390
commit 61831cb39155842c781f63214a7862bfd23e3390
Author: Peter Boström <pbos@chromium.org>
Date: Wed May 22 21:32:53 2024 +0000
Reland "Make NOTREACHED() without parameters [[noreturn]]"
This is a reland of commit fdeac78ccd92e2cc578d8cfa04219599520138b4
Original change's description:
> Make NOTREACHED() without parameters [[noreturn]]
>
> Previous callers have been migrated to NOTREACHED_IN_MIGRATION(). This
> also renames the few NOTREACHED_NORETURN() callers in base/ to
> NOTREACHED() to verify that it's still working (no missing-return).
>
> Let's see how many times this needs to be reverted and relanded. :)
>
> Bug: 40580068
> Change-Id: Ib8155e461ea9d0853c9508e28fc45389114da48c
> Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/5548908
> Code-Coverage: findit-for-me@appspot.gserviceaccount.com <findit-for-me@appspot.gserviceaccount.com>
> Commit-Queue: Lei Zhang <thestig@chromium.org>
> Owners-Override: Lei Zhang <thestig@chromium.org>
> Auto-Submit: Peter Boström <pbos@chromium.org>
> Commit-Queue: Peter Boström <pbos@chromium.org>
> Reviewed-by: Lei Zhang <thestig@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#1303439}
Bug: 40580068
This change replaces all uses of `absl::optional` with `std::*` variants
for optional. This is in line with upstream recent changes making
`absl::optional` a `typedef` to the `std` type.
This change has been done using an automated script:
#!/bin/bash
function replace {
echo "Replacing $1 by $2"
git grep -l "$1" \
| cut -f1 -d: \
| sort \
| uniq \
| grep \
-e "\.h" \
-e "\.cc" \
-e "\.mm" \
-e "\.py" \
| xargs sed -i "s/$1/$2/g"
}
function delete_line_with {
echo "Deleting lines with $1"
git grep -l "$1" \
| cut -f1 -d: \
| sort \
| uniq \
| grep \
-e "\.h" \
-e "\.cc" \
-e "\.mm" \
-e "\.py" \
| xargs sed -i "/$1/d"
}
function add_header {
echo "Adding header $1"
git diff --name-only HEAD \
| xargs ../tools/add_header.py --header "$1"
}
replace "absl::make_optional" "std::make_optional"
replace "absl::optional" "std::optional"
replace "absl::nullopt" "std::nullopt"
replace "absl::in_place" "std::in_place"
replace "absl::in_place_t" "std::in_place_t"
add_header "<optional>"
delete_line_with "\"third_party\/abseil-cpp\/absl\/types\/optional.h\""
git cl format
Chromium change:
https://chromium.googlesource.com/chromium/src/+/d9d21aa16829a7d471a4f3b3a493a31170ed8271
commit d9d21aa16829a7d471a4f3b3a493a31170ed8271
Author: David Benjamin <davidben@chromium.org>
Date: Mon Oct 2 23:29:57 2023 +0000
Make absl::optional a typedef for std::optional
This only changes the types around. It doesn't rewrite existing uses
to std::optional, which we can do incrementally.
absl::optional to std::optional seems to have two visible impacts.
First, the field order is different (bool first vs bool last).
std::optional's order (bool last) seems to be better overall, decreasing
binary size. Second, absl::optional's assertions crash with
__builtin_trap, while std::optional calls __libcpp_verbose_abort which
calls base::ImmediateCrash. __builtin_trap permits the compiler to
combine crash sites within a function but leads to worse crash
debugging. In base::ImmediateCrash, we'd made a conscious decision to
prefer debuggability and pay some binary size for it. The net size
increase brings our optional type in line with that preference.
For more details see the discussion and document below:
https://groups.google.com/a/chromium.org/g/cxx/c/XG3G85_ZF1k/m/_QN8adIJBQAJhttps://docs.google.com/document/d/1AW7q9HCLOk738OCj8Z2U_AKVUC0YIFZWuyRvv09XTHk/edit
Binary-Size: See discussion above.
Fuchsia-Binary-Size: See discussion above.
Bug: 1373619
cdesouza-chromium