This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH, committed] Support lambda templates.

On 13.09.2013 07:17, Adam Butcher wrote:
On 12.09.2013 23:44, Andrew MacLeod wrote:
On 09/12/2013 06:41 PM, Andrew MacLeod wrote:
On 09/12/2013 05:55 PM, Tobias Burnus wrote:
Adam Butcher wrote:
From: abutcher <abutcher@138bc75d-0d04-0410-961f-82ee72b054a4>
* lambda.c (lambda_function): Return template result if call operator is
    a template.

I believe that that patch causes the following build failure of cp/lambda.c:

In file included from ../../gcc/tree-core.h:27:0,
                 from ../../gcc/tree.h:23,
                 from ../../gcc/cp/lambda.c:27:
../../gcc/vec.h: In function 'void maybe_add_lambda_conv_op(tree)': ../../gcc/vec.h:626:44: error: 'direct_argvec' may be used uninitialized in this function [-Werror=maybe-uninitialized]
 vec_safe_space (const vec<T, A, vl_embed> *v, unsigned nelems)
../../gcc/cp/lambda.c:795:21: note: 'direct_argvec' was declared here
   vec<tree, va_gc> *direct_argvec;


PS: I am not positive since I use a patched tree and I haven't bisected.
I have verified that is the patch set causing the same issue for me. I backed out that revision and everything is fine.

Only locally of course to verify... :-P

Sorry about this.  I had built without -Werror.  I believed this
warning to be a false-positive with the maybe-uninitialized code which
I thought might be fixed later.  I will commit a fix to initialize
these pointers to 0.

Fixed in 202554.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]