Latest in magic cflags

[curaga]

http://lists.busybox.net/pipermail/busybox/2012-May/077816.html

On the busybox list it was pointed out that recent GCC includes a huge eh_frame section in binaries. Disabling this prevents debugging and pthread_cancel, but in a quick test cut 12% off a busybox build with our config.

So the magic is "-fno-asynchronous-unwind-tables". Adding it to the wiki, but announcing here too so everyone's aware

[AmatCoder]

recent GCC means 4.7.0 version, isn't?

[curaga]

No, I got such a section even with gcc 4.2. Not sure exactly what's the version cutoff here.

[bmarkus]

Just tried with tumbler, but got no any size rediction.

[curaga]

size -A /path/to/bin/tumbler | grep eh_frame

Before and after?

[AmatCoder]

No, I got such a section even with gcc 4.2. Not sure exactly what's the version cutoff here.

Since gcc 4.2? It does not seem possible....There must be a problem in busybox build system.

[AmatCoder]

I tried busybox 1.20.0 with gcc 4.6.1 without -fno-asynchronous-unwind-tables

Code: [Select]

tc@box:~/busybox-1.20.0$ make defconfig
tc@box:~/busybox-1.20.0$ make
tc@box:~/busybox-1.20.0$ size -A busybox | grep eh_frame
.eh_frame_hdr        44   135260064
.eh_frame           200   135260108
tc@box:~/busybox-1.20.0$ size -A busybox_unstripped | grep eh_frame
.eh_frame_hdr        44   135260064
.eh_frame           200   135260108

I don't get a huge .eh_frame section here... Is it produced by some custom configuration?

[curaga]

That I don't know. Our busybox config is available under release/src, though.

[AmatCoder]

I can not reproduce with /release/src/busybox_1.19.4_030812.tgz

Applied patches and config and I get a normal .eh_frame section... any further step?
Even binary included in archive has normal .eh_frame size...