summary refs log tree commit diff stats
path: root/results/scraper/launchpad/1347555
blob: 1b32a710319f62da6972e1c169cee165ee5b1631 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
qemu build failure, hxtool is a bash script, not a /bin/sh script

hxtool (part of the early build process) is a bash script.  Running it with /bin/sh yields a syntax error on line 10:

 10             STEXI*|ETEXI*|SQMP*|EQMP*) flag=$(($flag^1))

$(( expr )) is a bash extension, not part of /bin/sh.

Note that replacing the sh in the first line in hxtool with /bin/bash does not help, because the script is run manually from the Makefile with sh:

154         $(call quiet-command,sh $(SRC_PATH)/scripts/hxtool -h < $< > $@,"  GEN   $@")

The fix is to change those lines to

154         $(call quiet-command,bash $(SRC_PATH)/scripts/hxtool -h < $< > $@,"  GEN   $@")

(there are five or so).

On 07/23/2014 04:21 AM, Felix von Leitner wrote:
> Public bug reported:
> 
> hxtool (part of the early build process) is a bash script.  Running it
> with /bin/sh yields a syntax error on line 10:
> 
>  10             STEXI*|ETEXI*|SQMP*|EQMP*) flag=$(($flag^1))
> 
> $(( expr )) is a bash extension, not part of /bin/sh.

Wrong.  $(( expr )) is mandated by POSIX.  What system are you on where
/bin/sh is not POSIX?  (Solaris is the only platform where /bin/sh does
not try to be POSIX-compliant, but who uses that for qemu?)

What is the actual syntax error you are seeing?  Is this a bug in dash
on your distribution?  I can't get dash to fail for me on Fedora:

$ dash -c 'f=1; f=$(($f^1)); echo $f'
0
$ dash -n scripts/hxtool; echo $?
0

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org



I actually have bash installed as /bin/sh and /bin/bash.
But I also have heirloom sh installed, which installs itself as /sbin/sh, and that happened to be first in my $PATH.

Since the makefiles use "sh script" to run the scripts, that called the heirloom sh.

http://heirloom.sourceforge.net/sh.html

It is, it turns out, derived from OpenSolaris.  So there you go :-)

When I delete /sbin/sh, qemu builds.

On 07/23/2014 10:13 AM, Felix von Leitner wrote:
> I actually have bash installed as /bin/sh and /bin/bash.
> But I also have heirloom sh installed, which installs itself as /sbin/sh, and that happened to be first in my $PATH.
> 
> Since the makefiles use "sh script" to run the scripts, that called the
> heirloom sh.
> 
> http://heirloom.sourceforge.net/sh.html
> 
> It is, it turns out, derived from OpenSolaris.  So there you go :-)
> 
> When I delete /sbin/sh, qemu builds.

Then the bug is not in qemu, but in your environment.  Installing
known-broken heirloom where it can be found first on a PATH search for
sh is just asking for problems, not just with qemu, but with all SORTS
of programs that expect POSIX semantics from a Linux /bin/sh.

Rather than change the Makefile to invoke the script with bash, we could
instead bend over backwards to rewrite the script in a way that works
with non-POSIX shells (as in, flag=`expr $flag ^ 1`), but that feels
backwards to me.  Until someone is actively worried about porting qemu
to a true Solaris environment, rather than just an heirloom-as-/bin/sh
Linux environment, I don't think it's worth the effort.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org



On 23 July 2014 17:31, Eric Blake <email address hidden> wrote:
> Rather than change the Makefile to invoke the script with bash, we could
> instead bend over backwards to rewrite the script in a way that works
> with non-POSIX shells (as in, flag=`expr $flag ^ 1`), but that feels
> backwards to me.  Until someone is actively worried about porting qemu
> to a true Solaris environment, rather than just an heirloom-as-/bin/sh
> Linux environment, I don't think it's worth the effort.

My view on this has always been "we shouldn't assume bash,
but we can assume POSIX shell semantics". (And also that
we should assume /bin/sh is a POSIX shell, because it's the
21st century, and Solaris should just get with it :-))

thanks
-- PMM


It turns out that expr does not support ^ (at least according to the man page). :-)

Still, you could do expr -$flag + 1 to do the same thing.

Is the ruckus just about this one place where $(( )) is used or are there other non-Bourne-shell constructs?

Closing this ticket, as it was rather a problem with the non-posix-compliant shell and not the QEMU build system.