Re: Problem with recent tar on NetBSD/sparc-1.5

[Paul Eggert]

> Date: Sun, 9 Sep 2001 01:13:18 +0200 (CEST)
> From: Adrian Bunk <address@hidden>

>   1488 tar      CALL  wait4(0x5d1,0xeffff5f4,0,0)
>   1488 tar      RET   wait4 1489/0x5d1
>   1488 tar      CALL  getuid
>   1488 tar      RET   getuid 30313/0x7669
>   1488 tar      CALL  geteuid
>   1488 tar      RET   geteuid 30313/0x7669
>   1488 tar      CALL  getgid
>   1488 tar      RET   getgid 30313/0x7669
>   1488 tar      CALL  getegid
>   1488 tar      RET   getegid 30313/0x7669
>   1488 tar      CALL  write(0x2,0xefffed28,0x5)
>   1488 tar      GIO   fd 2 wrote 5 bytes
>        "tar: "
>   1488 tar      RET   write 5
>   1488 tar      CALL  write(0x2,0xefffed90,0x19)
>   1488 tar      GIO   fd 2 wrote 25 bytes
>        "Child died with signal 13"

It sounds like gzip is dying with signal 13, which I assume is
SIGPIPE.  It used to be that tar ignored all failures of the
underlying gzip process, but starting with 1.13.18 it started
reporting them, and that must be what you're running into.

Can you please go back a little further in the kdump output, and find
the area around the last successful read by tar from the gzip pipe?

Also, can you send the number of bytes returned by the last 10 or so
successful reads from that pipe?

Also, please also look in the kdump output and see what gzip was doing
just before it died.

Thanks.