On Wed, 26 Jul 2017 14:56:19 +0100 Emil Velikov <emil.l.veli...@gmail.com> wrote:
> From: Emil Velikov <emil.veli...@collabora.com>
>
> The option is used to indicate how the code will be used - would it be a
> part of shared or static one.
>
> In the former case one needs to export the specific symbols, although
> normally people want to statically build the protocol code into their
> project.
>
> If the option is missing a warning is emitted, to point people and do
> the right thing.
>
> Signed-off-by: Emil Velikov <emil.veli...@collabora.com>
> ---
> src/scanner.c | 61
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++-----
> 1 file changed, 56 insertions(+), 5 deletions(-)
>
> diff --git a/src/scanner.c b/src/scanner.c
> index c345ed6..cc45b74 100644
> --- a/src/scanner.c
> +++ b/src/scanner.c
> @@ -57,6 +57,11 @@ enum side {
> SERVER,
> };
>
> +enum object_type {
> + SHARED,
> + STATIC,
> +};
Hi,
I could go with this as well, as long as we have some solution to the
original problem of conflicting interface names that Jonas is trying to
solve.
> +
> static int
> usage(int ret)
> {
> @@ -70,6 +75,11 @@ usage(int ret)
> fprintf(stderr, " -h, --help display this help and
> exit.\n"
> " -v, --version print the wayland
> library version that\n"
> " the scanner was built
> against.\n"
> + " -t, --object-type=[static,shared]\n"
> + " How is the resulting
> code going to be built/used\n"
> + " static - standalone
> static object, used internally\n"
> + " shared - shared
> library, to be used by multiple projects\n"
> + " Using static is
> highly recommened.\n"
> " -c, --include-core-only include the core
> version of the headers,\n"
> " that is e.g.
> wayland-client-core.h instead\n"
> " of
> wayland-client.h.\n");
> @@ -1712,9 +1722,11 @@ emit_messages(struct wl_list *message_list,
> printf("};\n\n");
> }
>
> +
> static void
> -emit_code(struct protocol *protocol)
> +emit_code(struct protocol *protocol, enum object_type obj_type)
> {
> + const char *symbol_visibility;
> struct interface *i, *next;
> struct wl_array types;
> char **p, *prev;
> @@ -1728,6 +1740,19 @@ emit_code(struct protocol *protocol)
> "#include <stdint.h>\n"
> "#include \"wayland-util.h\"\n\n");
>
> + /* When building a shared library symbols must be exported, otherwise
> + * we want to have the symbols hidden. */
> + if (obj_type == STATIC) {
> + symbol_visibility = "WL_PRIVATE";
> + printf("#if defined(__GNUC__) && __GNUC__ >= 4\n"
> + "#define WL_PRIVATE __attribute__
> ((visibility(\"hidden\")))\n"
> + "#else\n"
> + "#define WL_PRIVATE\n"
> + "#endif\n\n");
> + } else {
> + symbol_visibility = "WL_EXPORT";
> + }
I wonder... would it be any better to just replace 'WL_EXPORT' and
'extern' with tokens:
- LOCAL_INTERFACE_DECL for declarations of symbols that will also be
defined in the generated code file
- EXTERN_INTERFACE_DECL for declarations of symbols that the generated
files will need but the code file will not define, that is,
interfaces defined on other XML files.
- LOCAL_INTERFACE_DEF for symbol definitions in the generated code file.
The exported configuration would then be:
LOCAL_INTERFACE_DECL=extern
EXTERN_INTERFACE_DECL=extern
LOCAL_INTERFACE_DEF=WL_EXPORT
That would be far too flexible and no-one would use it right, right?
> +
> wl_array_init(&types);
> wl_list_for_each(i, &protocol->interface_list, link) {
> emit_types_forward_declarations(protocol, &i->request_list,
> &types);
> @@ -1757,10 +1782,10 @@ emit_code(struct protocol *protocol)
> emit_messages(&i->request_list, i, "requests");
> emit_messages(&i->event_list, i, "events");
>
> - printf("WL_EXPORT const struct wl_interface "
> + printf("%s const struct wl_interface "
> "%s_interface = {\n"
> "\t\"%s\", %d,\n",
> - i->name, i->name, i->version);
> + symbol_visibility, i->name, i->name, i->version);
>
> if (!wl_list_empty(&i->request_list))
> printf("\t%d, %s_requests,\n",
> @@ -1790,6 +1815,24 @@ free_protocol(struct protocol *protocol)
> free_description(protocol->description);
> }
>
> +static enum object_type
> +parse_obj_type(const char *obj_type_str)
> +{
> + if (!obj_type_str) {
> + fprintf(stderr, "Warning: --object-type is not specified,
> assuming shared.\n");
> + return SHARED;
> + }
> +
> + if (strcmp(obj_type_str, "static") == 0)
> + return STATIC;
> +
> + if (strcmp(obj_type_str, "shared") == 0)
> + return SHARED;
> +
> + fprintf(stderr, "Error: invalid object type string '%s'\n",
> obj_type_str);
> + usage(EXIT_FAILURE);
> +}
> +
> int main(int argc, char *argv[])
> {
> struct parse_context ctx;
> @@ -1802,6 +1845,8 @@ int main(int argc, char *argv[])
> bool core_headers = false;
> bool version = false;
> bool fail = false;
> + char *obj_type_str = NULL;
> + enum object_type obj_type;
> int opt;
> enum {
> CLIENT_HEADER,
> @@ -1812,12 +1857,13 @@ int main(int argc, char *argv[])
> static const struct option options[] = {
> { "help", no_argument, NULL, 'h' },
> { "version", no_argument, NULL, 'v' },
> + { "object-type", required_argument, NULL, 't' },
> { "include-core-only", no_argument, NULL, 'c' },
> { 0, 0, NULL, 0 }
> };
>
> while (1) {
> - opt = getopt_long(argc, argv, "hvc", options, NULL);
> + opt = getopt_long(argc, argv, "hvtc", options, NULL);
Should be "hvt:c" I think?
>
> if (opt == -1)
> break;
> @@ -1829,6 +1875,9 @@ int main(int argc, char *argv[])
> case 'v':
> version = true;
> break;
> + case 't':
> + obj_type_str = optarg;
> + break;
> case 'c':
> core_headers = true;
> break;
> @@ -1858,6 +1907,8 @@ int main(int argc, char *argv[])
> else
> usage(EXIT_FAILURE);
>
> + obj_type = parse_obj_type(obj_type_str);
> +
> if (argc == 3) {
> input_filename = argv[1];
> input = fopen(input_filename, "r");
> @@ -1937,7 +1988,7 @@ int main(int argc, char *argv[])
> emit_header(&protocol, SERVER);
> break;
> case CODE:
> - emit_code(&protocol);
> + emit_code(&protocol, obj_type);
> break;
> }
>
The patch looks pretty much correct to me, if we choose to go this way.
Thanks,
pq
pgpVXFSxMPiRu.pgp
Description: OpenPGP digital signature
_______________________________________________ wayland-devel mailing list wayland-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/wayland-devel
