[PATCH v2] [gdb] Add spell check pre-commit hook

[Tom de Vries]

On 11/12/24 14:06, Tom de Vries wrote:
> Doing gdb/contrib/spellcheck.sh --check can be slow on large files.
> 
> [ The assumption below is that .git/wikipedia-common-misspellings.txt has been
> downloaded and that .git/spell-check.pat1.$md5sum has been generated. ]
> 
> Say we add typo "upto" at the first line of gdb/doc/gdb.texinfo, we get a
> quick result:
> ...
> $ time ./gdb/contrib/spellcheck.sh --check gdb/doc/gdb.texinfo
> 
> real    0m0,155s
> user    0m0,012s
> sys     0m0,029s
> ...
> but if we add the typo to the last line instead, we get:
> ...
> $ time ./gdb/contrib/spellcheck.sh --check gdb/doc/gdb.texinfo
> 
> real    0m9,234s
> user    0m0,013s
> sys     0m0,028s
> ...
> and likewise if there's no typo.
> 
> Add a --staged option to gdb/contrib/spellcheck.sh that doesn't look at the
> entire file, but only at the staged changes:
> ...
> $ git add gdb/doc/gdb.texinfo
> $ time ./gdb/contrib/spellcheck.sh --staged gdb/doc/gdb.texinfo
> Spell check failed in /home/vries/gdb/binutils-gdb.git/gdb/doc/gdb.texinfo
> 
> real    0m0,218s
> user    0m0,040s
> sys     0m0,031s
> ...
> which speeds things up a bit.
> 
> Now that it's no longer unreasonably slow, add a call to
> gdb/contrib/spellcheck.sh --staged in .pre-commit-config.yaml.
> 

An update on this.

I'm still pursuing this, but I think there are a few things that need 
fixing first.

We need a mechanism to opt out (more fine-grained than 
SKIP=spell-check).  For instance, we currently get:
...
$ ./gdb/contrib/spellcheck.sh gdb/ada-lang.c
TODO: onot -> note, not
...
for:
...
   {"Oabs", "\"abs\"", UNOP_ABS},
   {"Onot", "\"not\"", UNOP_LOGICAL_NOT},
   {"Oadd", "\"+\"", UNOP_PLUS},
...

Likewise in opcodes/ and sim/ we run into "andd->and" triggering on 
architectures where that is a valid opcode.

Also, I think we better get the auto-correct in place rather than just 
giving a simple "exited with status 1" type of message.  The reason I 
didn't do this yet, is because that part is still slow.  So, that part 
needs speeding up.

I'm currently working on these two issues.

Thanks,
- Tom

> Tested on aarch64-linux.  Verified with shellcheck.
> ---
>   .pre-commit-config.yaml   |  7 +++++++
>   gdb/contrib/spellcheck.sh | 44 ++++++++++++++++++++++++++++++++++++++-
>   2 files changed, 50 insertions(+), 1 deletion(-)
> 
> diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
> index 87726aeb758..6be4328f8c1 100644
> --- a/.pre-commit-config.yaml
> +++ b/.pre-commit-config.yaml
> @@ -22,3 +22,10 @@ repos:
>       - id: isort
>         types_or: [file]
>         files: 'gdb/.*\.py(\.in)?$'
> +  - repo: local
> +    hooks:
> +    - id: spell-check
> +      name: spell check
> +      language: script
> +      entry: ./gdb/contrib/spellcheck.sh --staged
> +      files: ^(gdb|gdbsupport|gdbserver)/
> diff --git a/gdb/contrib/spellcheck.sh b/gdb/contrib/spellcheck.sh
> index 1b3e88e259b..61ef15ef83a 100755
> --- a/gdb/contrib/spellcheck.sh
> +++ b/gdb/contrib/spellcheck.sh
> @@ -121,6 +121,10 @@ parse_args ()
>   		check=true
>   		shift
>   		;;
> +	    " --staged ")
> +		staged=true
> +		shift
> +		;;
>   	    *)
>   		break
>   		;;
> @@ -357,6 +361,7 @@ main ()
>   {
>       declare -a unique_files
>       check=false
> +    staged=false
>       parse_args "$@"
>   
>       get_dictionary
> @@ -365,9 +370,46 @@ main ()
>       declare -a replacements
>       parse_dictionary
>   
> -    # Reduce set of files for sed to operate on.
>       local files_matching_words
>       declare -a files_matching_words
> +
> +    if $staged; then
> +	declare -a tmpfiles
> +	tmpfiles=()
> +	trap 'rm -f "${tmpfiles[@]}"' EXIT
> +
> +	local orgfiles
> +	declare -A orgfiles
> +
> +	for f in "${unique_files[@]}"; do
> +	    local tmp
> +	    tmp=$(mktemp)
> +	    tmpfiles=("${tmpfiles[@]}" "$tmp")
> +	    orgfiles[$tmp]="$f"
> +
> +	    git diff --staged --minimal "$f" \
> +		| tail -n +5 \
> +		| grep -E "^\+" \
> +		| sed 's/^\+//' \
> +		      > "$tmp"
> +	done
> +
> +	mapfile -t files_matching_words \
> +		< <(find_files_matching_words "${tmpfiles[@]}")
> +
> +	if [ ${#files_matching_words[@]} -eq 0 ]; then
> +	    ret=0
> +	else
> +	    for f in "${files_matching_words[@]}"; do
> +		echo "Spell check failed in ${orgfiles[$f]}"
> +	    done
> +	    ret=1
> +	fi
> +
> +	exit $ret
> +    fi
> +
> +    # Reduce set of files for sed to operate on.
>       mapfile -t files_matching_words \
>   	    < <(find_files_matching_words "${unique_files[@]}")
>   
> 
> base-commit: 3d6f66d8bb6c539bc89b55b1add85dc6f53f2011