Help with solaris testing

Help with solaris testing

[Pedro Alves]

Hi guys!

(Mark, Joel, CCing you, as you're the usual suspects when we
 talk about either BSD or (!linux && !windows) testing :-) )

I wonder if you could give me a little help with this.

I've setup a bunch of VMWare VMs (x86) to be able to test changes I'm
making to BSD targets.  I've got FreeBSD 6,7, OpenBSD 4.3, and
OpenSolaris 10 all set up.  I set up an NFS mount on the host,
so all VMs share the same source tree.  All cool, and great
for easy testing.  FreeBSD and OpenBSD are OK, but ...

... solaris is giving be the troubles.

When running the testsuite under Solaris, I get a bunch of failures
and they appear to be related to something translating "\r" -> "\r\n".
The eols end up being "\r\r\n" instead of "\r\n", and a lot of test
patterns don't expect that.

How do people test under solaris?  Is this perhaps a well known
issue with an easy well known fix?

Same results either sshing to the VM, or testing on a local
terminal inside the VM.

E.g.:

 expect: does "break main\r\r\nBreakpoint 1: file ../../../src/gdb/testsuite/gdb.base/call-ar-st.c, line 1195.\r\r\n(gdb) " (spawn_id exp10) match regular expression "Breakpoint 
[0-9]* at .*: file .*, line [0-9]+.\r\n\(gdb\) $"? no
 "Breakpoint [0-9]*: file .*, line [0-9]+.\r\n\(gdb\)
 "Breakpoint [0-9]* at .*\(gdb\) $"? no
 "Breakpoint [0-9]* \(.*\) pending.*\(gdb\) $"? no
 "Make breakpoint pending.*y or \[n\]. $"? no
 "\(gdb\) $"? yes
 expect: set expect_out(0,string) "(gdb) "
 expect: set expect_out(spawn_id) "exp10"
 expect: set expect_out(buffer) "break main\r\r\nBreakpoint 1: file ../../../src/gdb/testsuite/gdb.base/call-ar-st.c, line 1195.\r\r\n(gdb) "
 FAIL: gdb.base/call-ar-st.exp: setting breakpoint at main

This is a default setup of OpenSolaris 10, with the only extra
things being gcc, gmake, etc, to be able to build gdb.

I've tried a bunch of stty settings, but perhaps I missed
something.  Maybe dejagnu overrides the stty settings anyway,
I don't know.

Running gdb throught od -c shows that gdb is only outputting a single '\n':

 0000360   v   =   0   x   8   0   4   7   c   c   0   )       a   t
 0000400   .   .   /   .   .   /   s   r   c   /   g   d   b   /   g   d
 0000420   b   .   c   :   2   8  \n   2   8  \t           m   e   m   s

That was a bit of:
 Temporary breakpoint 1, main (argc=1, argv=0x8047cc0) at ../../src/gdb/gdb.c:28
 28        memset (&args, 0, sizeof args);

Any hints?

 pedro@opensolaris:~$ runtest --version
 WARNING: Couldn't find the global config file.
 Expect version is       5.42.1
 Tcl version is          8.4
 Framework version is    1.4.4

 pedro@opensolaris:~$ uname -a
 SunOS opensolaris 5.11 snv_86 i86pc i386 i86pc Solaris

-- 
Pedro Alves

Re: Help with solaris testing

[Mark Kettenis]

> From: Pedro Alves <pedro@codesourcery.com>
> Date: Tue, 29 Jul 2008 15:58:21 +0100
> 
> Hi guys!
> 
> (Mark, Joel, CCing you, as you're the usual suspects when we
>  talk about either BSD or (!linux && !windows) testing :-) )
> 
> I wonder if you could give me a little help with this.
> 
> I've setup a bunch of VMWare VMs (x86) to be able to test changes I'm
> making to BSD targets.  I've got FreeBSD 6,7, OpenBSD 4.3, and
> OpenSolaris 10 all set up.  I set up an NFS mount on the host,
> so all VMs share the same source tree.  All cool, and great
> for easy testing.  FreeBSD and OpenBSD are OK, but ...
> 
> ... solaris is giving be the troubles.
> 
> When running the testsuite under Solaris, I get a bunch of failures
> and they appear to be related to something translating "\r" -> "\r\n".
> The eols end up being "\r\r\n" instead of "\r\n", and a lot of test
> patterns don't expect that.
> 
> How do people test under solaris?  Is this perhaps a well known
> issue with an easy well known fix?

This is strange; I've tested GDB on Solaris in the past and never
encountered this problem.

What does stty -a say on the system?  Could it be that one of ocrnl or
onlret is set?

Re: Help with solaris testing

[Pedro Alves]

[-- Attachment #1: Type: text/plain, Size: 1994 bytes --]

Hi Mark, thanks, and sorry for not replying sooner.  Only now
I came back to the issue.

On Tuesday 29 July 2008 16:39:15, Mark Kettenis wrote:

> > When running the testsuite under Solaris, I get a bunch of failures
> > and they appear to be related to something translating "\r" -> "\r\n".
> > The eols end up being "\r\r\n" instead of "\r\n", and a lot of test
> > patterns don't expect that.
> >
> > How do people test under solaris?  Is this perhaps a well known
> > issue with an easy well known fix?
>
> This is strange; I've tested GDB on Solaris in the past and never
> encountered this problem.
>
> What does stty -a say on the system?  Could it be that one of ocrnl or
> onlret is set?

Hmm, not really.

 -bash-3.2$ /usr/bin/stty -a
 ...
 opost -olcuc onlcr -ocrnl -onocr -onlret -ofill -ofdel tab3

I just found a workaround.  Overriding the stty settings expect does on
the pty used in spawn's,

 proc default_gdb_start { } {
  ...
  global stty_init
  set stty_init "<the stty -a settings of my terminal>"

  ...
 }

... makes the issue go away.

This bit here http://www.cotse.com/dlf/man/expect/spawn.htm:

 "Internally, spawn uses a pty, initialized the same way as the user’s tty.   
 This is further initialized so that all settings are “sane” (according to   
 stty(1))."

The latter seems to indicate where the problem is:

On solaris:
 pedro@opensolaris:~/orlando/gdb/multi_process/build-solaris/gdb$ /usr/bin/stty
 speed 38400 baud; -parity
 rows = 34; columns = 118; ypixels = 0; xpixels = 0;
 swtch = <undef>;
 brkint -inpck -istrip icrnl -ixon ixoff imaxbel onlcr onocr tab3 echo echoe echok echonl echoctl echoke iexten

On linux:
 [pedro@orlando][~/gdb/multi_process/src/gdb/testsuite/gdb.base]>/bin/stty
 speed 38400 baud; line = 0;
 -brkint -imaxbel iutf8

Probably, passing -nottyinit to spawn would "fix" it too.

Anybody seen this before?  What's the correct fix for this?

-- 
Pedro Alves

[-- Attachment #2: fix_expect_sol.diff --]
[-- Type: text/x-diff, Size: 880 bytes --]

---
 gdb/testsuite/lib/gdb.exp |    4 ++++
 1 file changed, 4 insertions(+)

Index: src/gdb/testsuite/lib/gdb.exp
===================================================================
--- src.orig/gdb/testsuite/lib/gdb.exp	2008-08-05 14:03:59.000000000 +0100
+++ src/gdb/testsuite/lib/gdb.exp	2008-08-05 14:42:40.000000000 +0100
@@ -1154,6 +1154,10 @@ proc default_gdb_start { } {
     global gdb_prompt
     global timeout
     global gdb_spawn_id;
+    global stty_init
+
+#    set stty_init raw
+    set stty_init "-parenb -parodd cs8 -hupcl -cstopb cread -clocal -crtscts -ignbrk brkint ignpar -parmrk -inpck -istrip -inlcr -igncr icrnl -ixon ixoff -iuclc -ixany imaxbel opost -olcuc -ocrnl onlcr onocr -onlret -ofill -ofdel nl0 cr0 tab3 bs0 vt0 ff0 isig icanon iexten echo echoe echok echonl -noflsh -xcase -tostop -echoprt echoctl echoke"
 
     gdb_stop_suppressing_tests;
 

Re: Help with solaris testing

[Daniel Jacobowitz]

On Tue, Aug 05, 2008 at 02:46:00PM +0100, Pedro Alves wrote:
> I just found a workaround.  Overriding the stty settings expect does on
> the pty used in spawn's,
> 
>  proc default_gdb_start { } {
>   ...
>   global stty_init
>   set stty_init "<the stty -a settings of my terminal>"
> 
>   ...
>  }
> 
> ... makes the issue go away.

Could you binary search to find the stty option that's a problem?  The
default in expect is "stty sane"; the Linux man page for stty defines
what "sane" is equivalent to, but you should check the OpenSolaris man
page to see what it says.  Or run stty sane / stty -a in a pty.

-- 
Daniel Jacobowitz
CodeSourcery

Re: Help with solaris testing

[Mark Kettenis]

> Date: Tue, 5 Aug 2008 10:03:16 -0400
> From: Daniel Jacobowitz <drow@false.org>
> 
> On Tue, Aug 05, 2008 at 02:46:00PM +0100, Pedro Alves wrote:
> > I just found a workaround.  Overriding the stty settings expect does on
> > the pty used in spawn's,
> > 
> >  proc default_gdb_start { } {
> >   ...
> >   global stty_init
> >   set stty_init "<the stty -a settings of my terminal>"
> > 
> >   ...
> >  }
> > 
> > ... makes the issue go away.
> 
> Could you binary search to find the stty option that's a problem?  The
> default in expect is "stty sane"; the Linux man page for stty defines
> what "sane" is equivalent to, but you should check the OpenSolaris man
> page to see what it says.  Or run stty sane / stty -a in a pty.

On Solaris 10, I get:


$ stty
speed 38400 baud; 
rows = 24; columns = 80; ypixels = 316; xpixels = 484;
csdata ?
eucw 1:0:0:0, scrw 1:0:0:0
intr = ^c; quit = ^\; erase = ^?; kill = ^u;
eof = ^d; eol = <undef>; eol2 = <undef>; swtch = <undef>;
start = ^q; stop = ^s; susp = ^z; dsusp = ^y;
rprnt = ^r; flush = <undef>; werase = ^w; lnext = ^v;
-parenb -parodd cs8 -cstopb -hupcl cread -clocal -loblk -crtscts -crtsxoff -parext 
-ignbrk brkint -ignpar -parmrk -inpck -istrip -inlcr -igncr icrnl -iuclc 
ixon -ixany -ixoff -imaxbel 
isig icanon -xcase echo echoe echok -echonl -noflsh 
-tostop echoctl -echoprt echoke -defecho -flusho -pendin iexten 
opost -olcuc onlcr -ocrnl -onocr -onlret -ofill -ofdel tab3 


and "stty sane" changes that to:


speed 38400 baud; 
rows = 24; columns = 80; ypixels = 316; xpixels = 484;
csdata ?
eucw 1:0:0:0, scrw 1:0:0:0
intr = ^c; quit = ^\; erase = ^?; kill = ^u;
eof = ^d; eol = <undef>; eol2 = <undef>; swtch = <undef>;
start = ^q; stop = ^s; susp = ^z; dsusp = ^y;
rprnt = ^r; flush = <undef>; werase = ^w; lnext = ^v;
parenb -parodd cs7 -cstopb -hupcl cread -clocal -loblk -crtscts -crtsxoff -parext 
-ignbrk brkint ignpar -parmrk -inpck istrip -inlcr -igncr icrnl -iuclc 
ixon -ixany -ixoff imaxbel 
isig icanon -xcase echo echoe echok -echonl -noflsh 
-tostop echoctl -echoprt echoke -defecho -flusho -pendin iexten 
opost -olcuc onlcr -ocrnl -onocr -onlret -ofill -ofdel 


which seems to be sane enough for me.  But the OpenSolaris defaults that Pedro showed don't seem sane to me.

Re: Help with solaris testing

[Pedro Alves]

On Tuesday 05 August 2008 15:29:34, Mark Kettenis wrote:

> On Solaris 10, I get:
>
>
> $ stty
> speed 38400 baud;
> rows = 24; columns = 80; ypixels = 316; xpixels = 484;
> csdata ?
> eucw 1:0:0:0, scrw 1:0:0:0
> intr = ^c; quit = ^\; erase = ^?; kill = ^u;
> eof = ^d; eol = <undef>; eol2 = <undef>; swtch = <undef>;
> start = ^q; stop = ^s; susp = ^z; dsusp = ^y;
> rprnt = ^r; flush = <undef>; werase = ^w; lnext = ^v;
> -parenb -parodd cs8 -cstopb -hupcl cread -clocal -loblk -crtscts -crtsxoff
> -parext -ignbrk brkint -ignpar -parmrk -inpck -istrip -inlcr -igncr icrnl
> -iuclc ixon -ixany -ixoff -imaxbel
> isig icanon -xcase echo echoe echok -echonl -noflsh
> -tostop echoctl -echoprt echoke -defecho -flusho -pendin iexten
> opost -olcuc onlcr -ocrnl -onocr -onlret -ofill -ofdel tab3
>

On a clean login, I get,

pedro@opensolaris:~/orlando/gdb/multi_process/build-solaris/gdb$ stty -a
speed 38400 baud; rows 30; columns 80;
intr = ^C; quit = ^\; erase = ^?; kill = ^U; eof = ^D; eol = <undef>;
eol2 = <undef>; swtch = <undef>; start = ^Q; stop = ^S; susp = ^Z; dsusp = ^Y;
rprnt = ^R; werase = ^W; lnext = ^V; flush = ^O;
-parenb -parodd cs8 -hupcl -cstopb cread -clocal -crtscts
-ignbrk brkint -ignpar -parmrk -inpck -istrip -inlcr -igncr icrnl -ixon -ixoff
-iuclc -ixany -imaxbel
opost -olcuc -ocrnl onlcr -onocr -onlret -ofill -ofdel nl0 cr0 tab3 bs0 vt0 
ff0
isig icanon iexten echo echoe echok -echonl -noflsh -xcase -tostop -echoprt
echoctl echoke

>
> and "stty sane" changes that to:
>
>
> speed 38400 baud;
> rows = 24; columns = 80; ypixels = 316; xpixels = 484;
> csdata ?
> eucw 1:0:0:0, scrw 1:0:0:0
> intr = ^c; quit = ^\; erase = ^?; kill = ^u;
> eof = ^d; eol = <undef>; eol2 = <undef>; swtch = <undef>;
> start = ^q; stop = ^s; susp = ^z; dsusp = ^y;
> rprnt = ^r; flush = <undef>; werase = ^w; lnext = ^v;
> parenb -parodd cs7 -cstopb -hupcl cread -clocal -loblk -crtscts -crtsxoff
> -parext -ignbrk brkint ignpar -parmrk -inpck istrip -inlcr -igncr icrnl
> -iuclc ixon -ixany -ixoff imaxbel
> isig icanon -xcase echo echoe echok -echonl -noflsh
> -tostop echoctl -echoprt echoke -defecho -flusho -pendin iexten
> opost -olcuc onlcr -ocrnl -onocr -onlret -ofill -ofdel
>

And after "stty sane":

pedro@opensolaris:~/orlando/gdb/multi_process/build-solaris/gdb$ /usr/bin/stty
speed 38400 baud; -parity
rows = 30; columns = 80; ypixels = 0; xpixels = 0;
brkint -inpck -istrip icrnl -ixany imaxbel onlcr
echo echoe echok echoctl echoke iexten

pedro@opensolaris:~/orlando/gdb/multi_process/build-solaris/gdb$ /usr/bin/stty -a
speed 38400 baud;
rows = 30; columns = 80; ypixels = 0; xpixels = 0;
csdata ?
eucw 1:0:0:0, scrw 1:0:0:0
intr = ^c; quit = ^\; erase = ^?; kill = ^u;
eof = ^d; eol = <undef>; eol2 = <undef>; swtch = ^z;
start = ^q; stop = ^s; susp = ^z; dsusp = ^y;
rprnt = ^r; flush = ^o; werase = ^w; lnext = ^v;
-parenb -parodd cs8 -cstopb -hupcl 
cread -clocal -loblk -crtscts -crtsxoff -parext
-ignbrk brkint ignpar -parmrk -inpck -istrip -inlcr -igncr icrnl -iuclc
ixon -ixany -ixoff imaxbel
isig icanon -xcase echo echoe echok -echonl -noflsh
-tostop echoctl -echoprt echoke -defecho -flusho -pendin iexten
opost -olcuc onlcr -ocrnl -onocr -onlret -ofill -ofdel

> which seems to be sane enough for me.  But the OpenSolaris defaults that
> Pedro showed don't seem sane to me.

Sorry, I may have mixed up what I showed, because I only learned after
the fact that "stty" != "stty sane".  :-)

Here's a diff of what you see after stty sane, and what I see after stty sane.
The ">" version is yours:

$ /usr/bin/stty -a
> speed 38400 baud;
  speed 38400 baud;
> rows = 24; columns = 80; ypixels = 316; xpixels = 484;
  rows = 30; columns = 80; ypixels = 0; xpixels = 0;
> csdata ?
  csdata ?
> eucw 1:0:0:0, scrw 1:0:0:0
  eucw 1:0:0:0, scrw 1:0:0:0
> intr = ^c; quit = ^\; erase = ^?; kill = ^u;
  intr = ^c; quit = ^\; erase = ^?; kill = ^u;
> eof = ^d; eol = <undef>; eol2 = <undef>; swtch = <undef>;
  eof = ^d; eol = <undef>; eol2 = <undef>; swtch = ^z;
> start = ^q; stop = ^s; susp = ^z; dsusp = ^y;
  start = ^q; stop = ^s; susp = ^z; dsusp = ^y;
> rprnt = ^r; flush = <undef>; werase = ^w; lnext = ^v;
  rprnt = ^r; flush = ^o; werase = ^w; lnext = ^v;
                      ^^^
> parenb -parodd cs7 -cstopb -hupcl cread -clocal -loblk -crtscts -crtsxoff
 -parenb -parodd cs8 -cstopb -hupcl cread -clocal -loblk -crtscts -crtsxoff
                 ^^^
> -parext -ignbrk brkint ignpar -parmrk -inpck istrip -inlcr -igncr icrnl
  -parext -ignbrk brkint ignpar -parmrk -inpck -istrip -inlcr -igncr icrnl
                                                ^^^^
> -iuclc ixon -ixany -ixoff imaxbel
  -iuclc ixon -ixany -ixoff imaxbel
> isig icanon -xcase echo echoe echok -echonl -noflsh
  isig icanon -xcase echo echoe echok -echonl -noflsh
> -tostop echoctl -echoprt echoke -defecho -flusho -pendin iexten
  -tostop echoctl -echoprt echoke -defecho -flusho -pendin iexten
> opost -olcuc onlcr -ocrnl -onocr -onlret -ofill -ofdel
  opost -olcuc onlcr -ocrnl -onocr -onlret -ofill -ofdel

I then went through the all the options I had used in the workaround,
and identified that it is setting onocr that makes the "\r\r\n" go away.

I'm now working around this by replacing /usr/bin/stty by:

 #!/bin/sh

 if [ $# == 1 ] && [ x"$1" == xsane ]
 then
     exec /usr/bin/stty.org $* onocr
 else
     exec /usr/bin/stty.org $*
 fi

The gdb.mi tests all fail because all commands that GDB sends
are echoed back :-(.  If I remove the onocr, the GDB commands are
not output at all.  :-(
MI testing is not important for me, though, so I consider
it a victory that I'm able to test at all...

-- 
Pedro Alves