Today I ran across a bug in the latest CVS libltdl on SPARC Solaris 9
and Sun's Studio 11 compiler.
The -ldl library is necessary order to use dlopen() on Solaris. When
configured using GCC I see this in config.log:
configure:7428: gcc-4.2.0 -std=gnu99 -o conftest -O2 -g -Wall -Winline
-W -W
Applied to HEAD.
Index: ChangeLog
from Gary V. Vaughan <[EMAIL PROTECTED]>
Add a test to simulate a failure visible on systems which
need a library prefix like 'lib', such as BeOS. Fix a number
of bugs exposed by this test in order for it to pass.
Currently t
Applied to HEAD.
Index: ChangeLog
from Gary V. Vaughan <[EMAIL PROTECTED]>
Reenable compilation of libltdl with C++:
* libltdl/ltdl.c (tryall_dlopen): Rename the try variable to
attempt, and cast explicitly.
(advise_dup): Add explicit cast.
Index: l
Howdy Bob,
On Jun 24, 2007, at 12:22 PM, Bob Friesenhahn wrote:
Libtool has moved to a "Cathedral" model of development. I don't
think it was always that way. There was a time when libtool was
pretty rough around the edges and users did part of the testing,
with lots of subsequent bug rep
On Sun, 24 Jun 2007, Gary V. Vaughan wrote:
In an attempt to get 2.2 out the door, I'm making the minimal changes
necessary to close release blocker bugs in the style of the existing code.
Part of the reason 2.2 is several years late is from trying to make
everything perfect.
I agree, and t
Howdy Bob,
On Jun 24, 2007, at 11:42 AM, Bob Friesenhahn wrote:
On Sun, 24 Jun 2007, Ralf Wildenhues wrote:
If we need to use pointers to void for hiding in external headers,
then
so be it. But most of the time I don't see why this:
struct lt_foo; /* declaration only, no definition */
typ
On Sun, 24 Jun 2007, Ralf Wildenhues wrote:
If we need to use pointers to void for hiding in external headers, then
so be it. But most of the time I don't see why this:
struct lt_foo; /* declaration only, no definition */
typedef struct lt_foo *lt_foo_handle;
int lt_foofunc (lt_foo_handle);
* Gary V. Vaughan wrote on Sun, Jun 24, 2007 at 04:02:57PM CEST:
> On Jun 24, 2007, at 7:27 AM, Ralf Wildenhues wrote:
>
> >Please add necessary casts to this patch so compilation with C++
> >compiler is not broken again.
>
> Sounds like we need a test that tries to compile libltdl with C++.
Hallo Ralf,
Thanks for the review :)
On Jun 24, 2007, at 7:27 AM, Ralf Wildenhues wrote:
Review for the remaining stuff:
[ tryall_dlopen ]
-const lt_dlvtable *vtable = 0;
-lt_dlloader *loader = 0;
+lt_dlloader loader = lt_dlloader_next (0);
+const lt_dlvtable *loader_vtable;
Review for the remaining stuff:
[ tryall_dlopen ]
> -const lt_dlvtable *vtable = 0;
> -lt_dlloader *loader = 0;
> +lt_dlloader loader = lt_dlloader_next (0);
> +const lt_dlvtable *loader_vtable;
>
> -while ((loader = (lt_dlloader *) lt_dlloader_next (loader)))
> +do
This
While reviewing Gary's 330 patch, I noticed that the patch that
introduced lt_dladvise also introduced some memleaks:
- advise_taken in tryall_dlopen is malloced several times,
unnecessarily, and not freed in each code path. AFAICS the code can
be greatly simplified by dropping advise_taken an
* Gary V. Vaughan wrote on Sat, Jun 23, 2007 at 11:04:44PM CEST:
>
> I didn't explain myself very well. I meant to say that I have several
> patches in quilt format in the .patches directory of my local libtool
> working copy. Some of them are half-cocked, some of them pending
> review, some of
12 matches
Mail list logo
