1 |
On 01/14/10 14:54, Cyprien Nicolas wrote: |
2 |
> Hello lispers, |
3 |
> |
4 |
> I'm attaching to this mail two ebuild for the last releases of Bigloo |
5 |
> and Hop, I have mainly adapted the existing ebuilds. |
6 |
> |
7 |
Hello Cyprien, |
8 |
|
9 |
Thanks for providing these ebuilds, I'll make sure they are added to the |
10 |
overlay today. |
11 |
> Some notes, about the bigloo.ebuild: |
12 |
> |
13 |
> I have to had the threads flag, for dependencies, hop[threads] needs |
14 |
> bigloo[threads], which needs boehm-gc[threads]. That makes the hop |
15 |
> ebuild with three dependencies on hop, I don't know how to make it |
16 |
> clearer, or better. |
17 |
> |
18 |
There was an issue with a bundled copy of boegm-gc, I hope Mr Serrano |
19 |
took care of that :), otherwise some patching must be made for security |
20 |
reasons. |
21 |
> Then, I had to manually had -fno-omit-frame-pointer to Bigloo's CFLAGS, |
22 |
> for several reasons: |
23 |
> First, for compile-bee, part of the code may be compiled with -pg flag |
24 |
> (for gprof profiling), this option is incompatible with |
25 |
> -fomit-frame-pointer. |
26 |
> Next, at runtime, Bigloo uses the CFLAGS he got at compile time when |
27 |
> calling GCC on the produced C file. so bigloo -p will call gcc -pg (if |
28 |
> enabled), then raise a gcc error. |
29 |
> |
30 |
Yes, I've bumped up to this problem too. |
31 |
Strictly "ebuild-speaking", the best way to handle this is with the |
32 |
eclass flag-o-matic, removing -fomit-frame-pointer. Either way, you are |
33 |
right, that flag must be disabled. |
34 |
> |
35 |
> This ebuild has been ran on both x86 hardened and x86_64 hardened |
36 |
> systems, without any issue except the -pg thing I have bypassed. |
37 |
> |
38 |
Great, thanks for the testing. |
39 |
|
40 |
Regards, |
41 |
Tomas |
42 |
> |
43 |
> |
44 |
> Any comment are welcome |
45 |
> |
46 |
> |
47 |
> Regards, |
48 |
> |
49 |
> ,Cyprien |
50 |
> |
51 |
> |