On 8/4/2009 7:13 AM, Graham Murray wrote:
Remy Blank<remy.bl...@pobox.com> writes:
The whole issue seems to be handled quite strangely IMO. You would think
breaking Python for all ~x86 is a major offense...
It did not break for all ~x86. I have 2 systems both running ~x86, both
have emerged (but not made active) python-3.1, /usr/bin/python is a bash
script on both yet emerge works with no problems. The only non-standard
things are that I have unmasked gcc-4.4.1 and am using git 2.6.31-rc
kernels.
Aha. I think I've discovered the "problem". There was a change in the
kernel execl() call between 2.6.27 and 2.6.28:
--- linux-2.6.27-hardened-r3/fs/binfmt_script.c 2008-10-09
18:13:53.000000000 -0400
+++ linux-2.6.28-hardened/fs/binfmt_script.c 2008-12-24
18:26:37.000000000 -0500
@@ -22,14 +22,15 @@
char interp[BINPRM_BUF_SIZE];
int retval;
- if ((bprm->buf[0] != '#') || (bprm->buf[1] != '!') ||
(bprm->sh_bang))
+ if ((bprm->buf[0] != '#') || (bprm->buf[1] != '!') ||
+ (bprm->recursion_depth > BINPRM_MAX_RECURSION))
return -ENOEXEC;
/*
* This section does the #! interpretation.
* Sorta complicated, but hopefully it will work. -TYT
*/
- bprm->sh_bang = 1;
+ bprm->recursion_depth++;
allow_write_access(bprm->file);
fput(bprm->file);
bprm->file = NULL;
The kernel >= 2.6.28 now supports nesting up to 4 levels of script in
the #! lines, if I'm reading that right, whereas < 2.6.28 it only
supported 1 level. I'll go update the b.g.o entry and try upgrading my
kernel. Though I dunno what that means for Gentoo/FreeBSD.
--Mike
