Patch: suppress duplicate Fl_Printer::class_id

[James C]

Hello All,

Building commit 317f010 (two days ago) of dillo, I got a duplicate
symbol on OSX 10.6.8 with gcc 4.2.1. Here is the diff that suppresses
it.

----------------
$ hg diff
diff -r c09473fb7311 dw/fltkplatform.cc
--- a/dw/fltkplatform.cc Sat Sep 13 09:46:28 2014 +0000
+++ b/dw/fltkplatform.cc Sat Sep 13 23:49:26 2014 +1200
@@ -37,10 +37,10 @@
* any need for currently. This stub prevents that. */
class FL_EXPORT Fl_Printer : public Fl_Paged_Device {
public:
- static const char *class_id;
+// static const char *class_id;
Fl_Printer(void) {};
};
-const char *Fl_Printer::class_id = "Fl_Printer";
+//const char *Fl_Printer::class_id = "Fl_Printer";

/* Tooltips */
static Fl_Menu_Window *tt_window = NULL;
---------------

I have no idea what it will turn out that I've broken.

If having this as a pull request will help, then I will try to work out how.

Regards,
James.

[eocene]

Post by James C
Building commit 317f010 (two days ago) of dillo, I got a duplicate
symbol on OSX 10.6.8 with gcc 4.2.1. Here is the diff that suppresses
it.
----------------
$ hg diff
diff -r c09473fb7311 dw/fltkplatform.cc
--- a/dw/fltkplatform.cc Sat Sep 13 09:46:28 2014 +0000
+++ b/dw/fltkplatform.cc Sat Sep 13 23:49:26 2014 +1200
@@ -37,10 +37,10 @@
* any need for currently. This stub prevents that. */
class FL_EXPORT Fl_Printer : public Fl_Paged_Device {
- static const char *class_id;
+// static const char *class_id;
Fl_Printer(void) {};
};
-const char *Fl_Printer::class_id = "Fl_Printer";
+//const char *Fl_Printer::class_id = "Fl_Printer";

To refresh my memory of the magnitude of what printer stuff we pull in
without the stub, I commented out the stub -- and saw nothing to do with
Fl_Printer with nm.

I thought maybe this somehow had to do with the changes that Lauri Kasanen
made to fltk last month like
http://www.fltk.org/newsgroups.php?s10767+gfltk.commit+v10781
http://www.fltk.org/newsgroups.php?s10787+gfltk.commit+v10787
but I see that Manolo took out the Fl_Printer::class_id check in
Fl_Text_Display in March 2012. 1.3.2 came out in...let's see...
December 2012.

So we should probably remove the stub altogether.

[James C]

Removing the stub on OSX 10.6.8 with fltk 1.3.2 does not make dillo grow.

Sizes of dillo image; the first line counts shared libraries, and the
second overestimates, perhaps consistently, the size of the to-run
code:

- with Fl_Printer stub without class_id:
$ otool -L src/dillo | wc
14 92 1370
$ ls -l src/dillo
-rwxr-xr-x 1 ... staff 2407760 15 Sep 21:21 src/dillo

- without Fl_Printer stub:
$ otool -L src/dillo | wc
14 92 1370
$ ls -l src/dillo
-rwxr-xr-x 1 ... staff 2407760 17 Sep 00:05 src/dillo

If it's purpose has been overtaken by events, and it increases the
brittleness of the code, then I would be happy for it to be removed.

Post by eocene

Post by James C
Building commit 317f010 (two days ago) of dillo, I got a duplicate
symbol on OSX 10.6.8 with gcc 4.2.1. Here is the diff that suppresses
it.
----------------
$ hg diff
diff -r c09473fb7311 dw/fltkplatform.cc
--- a/dw/fltkplatform.cc Sat Sep 13 09:46:28 2014 +0000
+++ b/dw/fltkplatform.cc Sat Sep 13 23:49:26 2014 +1200
@@ -37,10 +37,10 @@
* any need for currently. This stub prevents that. */
class FL_EXPORT Fl_Printer : public Fl_Paged_Device {
- static const char *class_id;
+// static const char *class_id;
Fl_Printer(void) {};
};
-const char *Fl_Printer::class_id = "Fl_Printer";
+//const char *Fl_Printer::class_id = "Fl_Printer";

To refresh my memory of the magnitude of what printer stuff we pull in
without the stub, I commented out the stub -- and saw nothing to do with
Fl_Printer with nm.
I thought maybe this somehow had to do with the changes that Lauri Kasanen
made to fltk last month like
http://www.fltk.org/newsgroups.php?s10767+gfltk.commit+v10781
http://www.fltk.org/newsgroups.php?s10787+gfltk.commit+v10787
but I see that Manolo took out the Fl_Printer::class_id check in
Fl_Text_Display in March 2012. 1.3.2 came out in...let's see...
December 2012.
So we should probably remove the stub altogether.
_______________________________________________
Dillo-dev mailing list
http://lists.dillo.org/cgi-bin/mailman/listinfo/dillo-dev