Bugzilla Bug 71396

I get the following error when emerging the above package:

g++ details.o diefast.o direct.o filter.o freebsd.o globalvars.o gps.o history.o linuxprocfs.o msgbox.o netpolling.o netwatch.o pipedtop.o polling.o quickbar.o renice.o tabledata.o treeview.o tstring.o userbox.o userinfo.o watchp.o -o gps -L/usr/lib -L/usr/X11R6/lib -lgtk -lgdk -rdynamic -lgmodule -lglib -ldl -lXi -lXext -lX11 -lm    -I/usr/include/gtk-1.2 -I/usr/include/glib-1.2 -I/usr/lib/glib/include -I/usr/X11R6/include -DWDOC=\"/usr/doc\"
history.o(.text+0x9bc): In function `destroy_history(_GtkWidget*, void*)':
: undefined reference to `pthread_join'
netpolling.o(.text+0x457): In function `NetworkActor::hopen(char*, int)':
: undefined reference to `pthread_cancel'
netpolling.o(.text+0x46a): In function `NetworkActor::hopen(char*, int)':
: undefined reference to `pthread_join'
netpolling.o(.text+0x51a): In function `NetworkActor::hopen(char*, int)':
: undefined reference to `pthread_join'
collect2: ld returned 1 exit status

Happens with glibc-2.3.4.20040808-r1

seemant: you added this package... it has no metadata.xml

looks like the package just needs to compile with '-lpthread'

should be fixed in 1.1.0-r1.  Please test and report.

Emerges without problems now, but core dumps when started....

can you post me an strace log of it?

Hmmm, that's an interesting one.

GDB doesn't help much:
(gdb) where
#0  0x0805988c in refresh() ()
#1  0x08056a6e in main ()

Nor does strace (tail of about 700 lines):
shmctl(196613, IPC_64|IPC_RMID, 0)      = 0
write(3, "5\1\4\0M\0@\1\4\0@\0010\0009\0007\0\5\0N\0@\1M\0@\1\0\0"..., 676) = 67
read(3, "\1\0\342\2\0\0\0\0\4\0@\1\0\0\0\0\0\0\0\0\20\0\0\0008\t"..., 32) = 32
shmdt(0x27594000)                       = 0
write(3, "\213\2\2\0L\0@\1\24\0\6\0\4\0@\1#\0\0\0#\0\0\0\0\0\0\0"..., 32) = 32
read(3, "\1 \344\2\t\0\0\0#\0\0\0\0\0\0\0\t\0\0\0\0\0\0\0008\f%"..., 32) = 32
read(3, "C\0\0\0\1\0\0\0\1\0\0\0\266\367\"\'l\31\7X<\264\34\'\0"..., 36) = 36
getpriority(PRIO_PROCESS, 0)            = 20
setpriority(PRIO_PROCESS, 0, 8)         = 0
getpriority(PRIO_PROCESS, 0)            = 12
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
+++ killed by SIGSEGV (core dumped) +++

However, when I ./configure and make by hand as root in the portage working dir, I get a gps which *works*.

Manual ./configure says:
--------------------------------------------------------------------------
gPS will be compiled with the following polling modes:
(The default will be the first of this list)

* Linux /proc fs polling   (OS: Linux        , Architecture: Any)

--------------------------------------------------------------------------

During emerge, ./configure says:
--------------------------------------------------------------------------
gPS will be compiled with the following polling modes:
(The default will be the first of this list)

No polling methods have been found. When not
in directly supported systems (Linux/FreeBSD), the only
method available is top polling. As top polling has not
been acknowledged, this system may not have top or
have a not supported one. You can download a working
top from ftp://ftp.groupsys.com/pub/top

Compiling without a polling method makes no sense,
fix it before you compile.

--------------------------------------------------------------------------

After removing "userpriv" and "usersandbox" from my /etc/make.conf,
emerge also produces a working gps.

As I use a hardened kernel with most pax and grsec features enabled, I believe some ./configure tests give wrong results on my system when not run as root.

I'm on leave of absence for the next 45-90 days, sorry :/

Solar, suggestions on this please?

other than CFLAGS + = -pthread and to use setpriority() you need root privs..
Without a gdb full backtrace there is not much that I can tell from this
without looking at the source code in question. So sorry seemant not tot
really. 

dunno what to tell you Klaus

Hmmm, output a big redstarred message when emerge gps is run with the wrong
settings?
(just to remind me that I have to turn off sandboxing when emerging any future
upgrade to gps)

Brian, is there some way around this, other than turning off {user,}sandbox?

RESTRICT="nouserpriv" in the ebuild might do the trick.

Klaus, I've added the restrict command, as Brian suggested.

When you have a moment, can you resync (and make sure the RESTRICT line is in the ebuild), and re-emerge this?  Please do let me know how it turns out.

Works fine here.

thanks for bearing with me Klaus and solar and Brian :)