1 |
Hi, all, |
2 |
|
3 |
For some time I've been getting messages like: |
4 |
|
5 |
cannot properly execute |
6 |
/var/lib/layman/science/virtual/lapack/lapack-3.6-r100.ebuild |
7 |
|
8 |
for *every* package in the science overlay. This happened on 2 of 3 |
9 |
very similar gentoo |
10 |
systems I maintain, but not the third. As far as I could tell, the |
11 |
permissions and |
12 |
ownership of all /etc/portage and /usr/portage and /var/lib/layman |
13 |
files and directories |
14 |
are identical. All the machines have the same /etc/passwd and /etc/group files, |
15 |
the same /var/lib/portage/world files, and so on. I've gone to some |
16 |
effort to make these |
17 |
machines as identical as possible. I have done layman -d science then layman -a |
18 |
science to see if wiping out the overlay then reinstalling it would |
19 |
help, but that makes no |
20 |
difference. |
21 |
|
22 |
Googling the error produces just a few hits that recommend permissions |
23 |
and ownerships |
24 |
that helped solve this problem for others, but don't help me. |
25 |
|
26 |
It appears that this error message comes from /usr/bin/eix. I've |
27 |
looked through the |
28 |
source code a bit, but it's been long enough since I've messed with C |
29 |
that this didn't |
30 |
yield any answers. |
31 |
|
32 |
strace indicates that eix does an lstat on every element of |
33 |
/var/lib/layman/science/virtual/lapack/lapack-3.6-r100.ebuild before |
34 |
deciding to print the |
35 |
error message. It also indicates that eix does a setuid and setgid to |
36 |
portage:portage |
37 |
early in its execution. As far as I can tell, portage:portage ought |
38 |
to be able to both read |
39 |
and execute this ebuild (and all the other ones eix complains about). |
40 |
|
41 |
And now, out of the blue apparently, the third system has started |
42 |
spitting out this error |
43 |
message as well. |
44 |
|
45 |
Any ideas on how to debug this? |
46 |
|
47 |
John Blinka |