1 |
On 8/2/07, Sourav K. Mandal <sourav.mandal@×××××.com> wrote: |
2 |
> [quote snipped] |
3 |
> |
4 |
> Comments: |
5 |
> |
6 |
> * I would be nervous to use the ~x86 keyword except for specific |
7 |
> packages at a high-level. The reason it takes a while for things to get |
8 |
> into x86 is because there are obscure bugs which can affect obscure |
9 |
> setups -- unfortunately, GNUstep is in this category. A small bug in |
10 |
> the toolchain can screw up everything else on top of it. At a minimum, |
11 |
> I would revert to gcc-4.1 and glibc-2.5 in x86. |
12 |
I found too many problems related with having some packages ~x86 and |
13 |
some x86, and in general my experience with ACCEPT_KEYWORDS="~x86" |
14 |
has been a really pleasant one (I mean, I've been using gentoo for years now, |
15 |
and ~x86 is alot more stable now than it was before). |
16 |
I tried reverting to gcc-4.1.2 and recompiling gnustep-base but it didn't |
17 |
change anything, I can try reverting glibc as well, but I'd rather try other |
18 |
things first if possible. |
19 |
|
20 |
> |
21 |
> * Consider adding "-ggdb" to your C(XX)FLAGS, and "splitdebug" to your |
22 |
> features. Compilation will produce gdb-compatible debugging symbols, |
23 |
> but they'll be split off into /usr/lib/debug to reduce loading times for |
24 |
> binaries. This works for ObjC/GNUstep; if you are serious about |
25 |
> debugging, do this. |
26 |
I've tried this already, but the debug flags are being stripped so far |
27 |
as I can tell. Take |
28 |
a look at the example below which I randomly selected from the |
29 |
compilation output : |
30 |
|
31 |
i686-pc-linux-gnu-gcc NSIndexPath.m -c \ |
32 |
-MMD -MP -DGNUSTEP_TARGET_DIR=\".\" |
33 |
-DGNUSTEP_TARGET_CPU=\"ix86\" -DGNUSTEP_TARGET_OS=\"linux-gnu\" |
34 |
-DGNUSTEP_IS_FLATTENED=\"yes\" -DLIBRARY_COMBO=\"gnu-gnu-gnu\" -Wall |
35 |
-Wcast-align -DGNUSTEP -DGNUSTEP_BASE_LIBRARY=1 -DGNU_GUI_LIBRARY=1 |
36 |
-DGNU_RUNTIME=1 -DGNUSTEP_BASE_LIBRARY=1 -D_REENTRANT -fPIC -g -Wall |
37 |
-DDEBUG -fno-omit-frame-pointer -DGSWARN -DGSDIAGNOSE -Wno-import |
38 |
-march=prescott -pipe -fno-strict-aliasing -fexceptions |
39 |
-fobjc-exceptions -D_NATIVE_OBJC_EXCEPTIONS -fgnu-runtime |
40 |
-fconstant-string-class=NSConstantString -I../Headers/Additions |
41 |
-I../Headers -I./. -I. -I/usr/GNUstep/System/Library/Headers |
42 |
-I/usr/GNUstep/System/Library/Headers -I/usr/include/libxml2 |
43 |
-I/usr/include -I/var/tmp/portage/gnustep-base/gnustep-base-1.14.0/temp/GNUstep/Library/Headers |
44 |
-I/usr/GNUstep/Local/Library/Headers |
45 |
-I/usr/GNUstep/System/Library/Headers \ |
46 |
-o obj/NSIndexPath.o |
47 |
|
48 |
This is with -ggdb on the CFLAGS, splitdebug on FEATURES and debug on |
49 |
gnustep-base's |
50 |
USEflags. Am I doing anything wrong ? |
51 |
|
52 |
|
53 |
> |
54 |
> * You have a ton of USE flags. The more intricate/featureful your |
55 |
> system, the more likely you are to get weird bugs. Consider paring it |
56 |
> down to exactly what you need. |
57 |
Ok, there are lots of USE flags there that are related to a bug on |
58 |
gtkphoto2 which |
59 |
I still haven't had the time to take a better look at (I've got a |
60 |
canon camera, but the |
61 |
canon useflag is not adding support for it, so I brute forced and set |
62 |
all useflags |
63 |
possible). I can cut down my USEflags, but that would involve rebuilding the |
64 |
entire system, and I would really prefer another solution to finding |
65 |
out what's wrong |
66 |
if possible. |
67 |
I mean, my idea with building gnustep is basically to test Étoilé for some time. |
68 |
|
69 |
cheers, |
70 |
--lf |
71 |
|
72 |
> |
73 |
> |
74 |
> Good luck, |
75 |
> |
76 |
> Sourav |
77 |
> |
78 |
> |
79 |
> -- |
80 |
> Sourav K. Mandal |
81 |
> http://sourav.net/ |
82 |
> |
83 |
> PGP: 7E7E 14CD A983 484C 8A43 55CA DBAC 539C 1814 3DAF |
84 |
> |
85 |
> -- |
86 |
> gentoo-gnustep@g.o mailing list |
87 |
> |
88 |
> |
89 |
|
90 |
|
91 |
-- |
92 |
"People assume that time is a strict progression of cause to effect... |
93 |
but actually, from a non-linear, non-subjective viewpoint, it's more |
94 |
like a big ball of wibbly-wobbly...timey-wimey...stuff." - The Doctor |
95 |
-- |
96 |
gentoo-gnustep@g.o mailing list |