-
Notifications
You must be signed in to change notification settings - Fork 431
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fedora 20 / gcc 4.8.3 compilation is broken #23
Comments
The nice way to do it is introduce GCC version check like here: http://stackoverflow.com/questions/7067385/find-the-gcc-version For 4.8.3 we may remove the warning and keep it for all the rest |
Compile broke on our whale system with gcc 4.6.2 and 4.9.1 as well. Why not remove the variable/assignment and keep the stricter error checking? tony |
The variable is actually used. |
I can see it being assigned to (declared #305, assigned #343), but no other reference to that variable. Is this assignment affecting line #273 too? tony |
As far as I remember, It is passed somewhere to the rte_call() |
On Nov 13, 2014, at 9:39 PM, Pasha <[email protected]mailto:[email protected]> wrote: As far as I remember, It is passed somewhere to the rte_call() Could we then just do the following to set the persistent group_index value? diff --git a/test/perf/libperf.c b/test/perf/libperf.c ucs_status_t uct_perf_test_setup_endpoints(uct_perf_context_t *perf)
address = (uintptr_t)perf->super.recv_buffer;
perf->peers = calloc(group_size, sizeof(*perf->peers)); |
I think you are actually right. |
On Nov 13, 2014, at 10:13 PM, Pasha <[email protected]mailto:[email protected]> wrote: I think you are actually right. Had to happen sooner or later :) group_index = rte_call(&perf->super, group_index); is not used anywhere and can be removed. Can you please submit pull request. Mellanox testing will pickup it automatically. I wonder if rte_call has a side-effect that is being picked up on line 273 but haven’t disentangled the pointer/struct chain yet -- calling the local variable “group_index” as well might be a bit of a red herring. But there are subsequent compile issues so haven’t been able to verify (will do separate report to keep this one clean). tony |
On Nov 13, 2014, at 10:28 PM, Curtis, Tony <[email protected]mailto:[email protected]> wrote: group_index = rte_call(&perf->super, group_index); is not used anywhere and can be removed. Can you please submit pull request. Mellanox testing will pickup it automatically. I wonder if rte_call has a side-effect that is being picked up on line 273 but haven’t disentangled the pointer/struct chain yet -- calling the local variable “group_index” as well might be a bit of a red herring. test/perf/perftest.c makes me think the group_index() call is simply meant to be pure and return a value, in which case we can simply delete the variable and assignment. I would like a second pair of more-experience-with-this-code eyes on it before doing anything… tony |
yes, |
…req-send-lane UCP/STREAM: Set correct req->send.lane in stream/short progress
Nothing wrong with the code. I think gcc 4.8.3 is broken.
Workaround for this:
So, I suggest remove
-Werror
. Any other ideas ?The text was updated successfully, but these errors were encountered: