You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Before filing a bug, please check whether it has been fixed since
the latest release: run "hg pull -u" and retry what you did to
reproduce the problem. Thanks.
What steps will reproduce the problem?
1. Build any go program, e.g. hello world
2. Strip it, with or with out strip -d
3. Run it and watch it commit suicide
What is the expected output? What do you see instead?
$ strip -d hello
BFD: hello: warning: Empty loadable segment detected, is this intentional ?
$ ./hello
SIGSEGV: segmentation violation
Faulting address: 0x99000000
PC=0x3b
Segmentation fault: 11 (core dumped)
What is your $GOOS? $GOARCH?
FreeBSD/386
Which revision are you sync'ed to? (hg log -l 1)
changeset: 4134:3c97caf16ae0
Please provide any additional information below.
Strip probably doesn't need to work, but it shouldn't
cause the binary to shoot itself in the head.
Changed since recent changes, but still fails (same on i386, but this is an issue on
amd64 too):
> strip -d 6.out
BFD: 6.out: warning: Empty loadable segment detected, is this intentional ?
BFD: stE2SbKS: warning: allocated section `.interp' not in segment
> ./6.out
ELF interpreter ^?ELF^B^A^A not found
Abort
Does this happen on Linux?
This is a known issue. If you strip the binary you throw
away the runtime symbol information that gets used for
stack traces and some aspects of reflection. Don't do that. ;-)
We plan to move that data into a rodata segment and then
strip will not break the binary but will also not remove
very much.
Activity
dhobsd commentedon Nov 19, 2009
Comment 1:
rsc commentedon Nov 19, 2009
Comment 2:
Status changed to LongTerm.
gopherbot commentedon Nov 19, 2009
Comment 3 by wjosephson:
dhobsd commentedon Nov 20, 2009
Comment 4:
Where do the runtime symbols get stored? Everywhere that I'm seeing right now implies that they're either in .hash or .gosymtab which would suggest that if those sections are read-only, they shouldn't get tossed by strip. But they are: > objdump -hrt ../gofmt/gofmt ../gofmt/gofmt: file format elf64-x86-64 Sections: Idx Name Size VMA LMA File off Algn 0 .interp 00000015 00000000004007eb 00000000004007eb 000007eb 2**0 CONTENTS, ALLOC, LOAD, READONLY, DATA 1 .got 00000000 0000000000446678 0000000000446678 00046678 2**3 CONTENTS, ALLOC, LOAD, DATA 2 .got.plt 00000000 0000000000446d48 0000000000446d48 00046d48 2**3 CONTENTS, ALLOC, LOAD, DATA 3 .dynsym 00000018 0000000000464a28 0000000000464a28 00064a28 2**3 CONTENTS, ALLOC, LOAD, READONLY, DATA 4 .dynstr 00000004 0000000000446470 0000000000446470 00046470 2**0 CONTENTS, ALLOC, LOAD, READONLY, DATA 5 .hash 00000010 0000000000492ae0 0000000000492ae0 00092ae0 2**3 CONTENTS, ALLOC, LOAD, READONLY, DATA 6 .rela 00000000 00000000004472d8 00000000004472d8 000472d8 2**3 CONTENTS, ALLOC, LOAD, READONLY, DATA 7 .dynamic 00000090 0000000000482e48 0000000000482e48 00082e48 2**3 CONTENTS, ALLOC, LOAD, DATA 8 .text 00044f26 0000000000400800 0000000000400800 00000800 2**3 CONTENTS, ALLOC, LOAD, READONLY, CODE 9 .data 00060748 0000000000446000 0000000000446000 00046000 2**3 CONTENTS, ALLOC, LOAD, DATA 10 .bss 002145d0 00000000004a6748 00000000004a6748 000a6748 2**3 ALLOC 11 .gosymtab 0007664f 0000000000000000 0000000000000000 000a7008 2**0 CONTENTS, READONLY 12 .gopclntab 000057e6 0000000000000000 0000000000000000 0011d657 2**0 CONTENTS, READONLY SYMBOL TABLE: no symbolsrsc commentedon Apr 28, 2010
Comment 5:
This issue was closed by revision 000ab98.
Status changed to Fixed.