Ioan-Adrian Ratiu <adi@adirat.com> writes:

>      notmuch_status_t status;
> +    unsigned long filesize;
> +    char *filesize_str;
>  
>      status = _notmuch_message_file_get_mime_message (message_file,
>  						     &mime_message);
> @@ -464,6 +466,14 @@ _notmuch_message_index_file (notmuch_message_t *message,
>      subject = g_mime_message_get_subject (mime_message);
>      _notmuch_message_gen_terms (message, "subject", subject);
>  
> +    filesize = _notmuch_message_file_get_size (message_file);
> +    filesize_str = talloc_asprintf(NULL, "%lu", filesize);
> +    if (! filesize_str)
> +	return NOTMUCH_STATUS_OUT_OF_MEMORY;
> +
> +    _notmuch_message_add_term (message, "filesize", filesize_str);
> +    talloc_free (filesize_str);

I wanted to see what the performance impact of this commit was, but
unfortunately this breaks "notmuch new" (and most other things) as the
prefix "filesize" is not defined yet.

A handy way to see that this doesn't happen is to run

  % git rebase -i --exec "make clean test" origin/master

before posting a series.

Another preliminary question is why are you adding both a prefixed term
and a value slot for file size? You can use the value slot to query.

d