On Tue, Feb 13, 2024 at 08:18:57PM +0100, Omar Polo wrote:
> both update_timestamps may be used un-initalized.

You mean in the case where status is something other than
add/modify/delete and S_ISLNK(te->mode)) is true and ie is NULL?
I am not sure when this combination of conditions will happen in practice.
But initializing to 0 is probably the right thing to do in any case.

> Where here, do what I
> should have done in ef6234452a and remove the nested update_timestamps.

Yes, makes sense.

ok

> spotted by smatch.
> 
> ok?
> 
> diff -s /home/op/w/got
> commit - 06bd8ee415aa89adaeb49825ad96dfe83554b77e
> path + /home/op/w/got (staged changes)
> blob - 866c19c894b29dffe0ffd1be3cbc2f6999ed8db4
> blob + fc674d163162822cb326fcd1144a79b40778c452
> --- lib/worktree.c
> +++ lib/worktree.c
> @@ -1920,7 +1920,7 @@ update_blob(struct got_worktree *worktree,
>  	unsigned char status = GOT_STATUS_NO_CHANGE;
>  	struct stat sb;
>  	int fd1 = -1, fd2 = -1;
> -	int update_timestamps;
> +	int update_timestamps = 0;
>  
>  	if (asprintf(&ondisk_path, "%s/%s", worktree->root_path, path) == -1)
>  		return got_error_from_errno("asprintf");
> @@ -2044,7 +2044,6 @@ update_blob(struct got_worktree *worktree,
>  		goto done;
>  
>  	if (status == GOT_STATUS_MODIFY || status == GOT_STATUS_ADD) {
> -		int update_timestamps;
>  		struct got_blob_object *blob2 = NULL;
>  		char *label_orig = NULL;
>  		if (got_fileindex_entry_has_blob(ie)) {
> 
>