[PATCH, testsuite] Avoid testsuite harness failure on i686-mingw32 host

[PATCH, testsuite] Avoid testsuite harness failure on i686-mingw32 host

[Maxim Kuvyrkov]

[-- Attachment #1: Type: text/plain, Size: 357 bytes --]

This patch improves testing on i686-mingw32 host.  One of the tests in interrupt.exp assumes that Control-D is EOF, which is not the case for Windows hosts.  After the test timeouts the testsuite doesn't get GDB prompt back, which causes all subsequent tests to fail.

OK to apply?

Thank you,

--
Maxim Kuvyrkov
CodeSourcery / Mentor Graphics



[-- Attachment #2: gdb-win32-eof.ChangeLog --]
[-- Type: application/octet-stream, Size: 150 bytes --]

2012-11-15  Maxim Kuvyrkov  <maxim@codesourcery.com>

	gdb/testsuite/
	* gdb.base/interrupt.exp: Workaround problem of sending EOF on Windows
	hosts.

[-- Attachment #3: gdb-win32-eof.patch --]
[-- Type: application/octet-stream, Size: 1492 bytes --]

Index: gdb/testsuite/gdb.base/interrupt.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/interrupt.exp,v
retrieving revision 1.22
diff -u -p -r1.22 interrupt.exp
--- gdb/testsuite/gdb.base/interrupt.exp	16 Jan 2012 16:21:44 -0000	1.22
+++ gdb/testsuite/gdb.base/interrupt.exp	16 Nov 2012 03:43:06 -0000
@@ -194,14 +194,28 @@ if ![file exists $binfile] then {
 	    }
 	}
 
-	send_gdb "\004"
-	gdb_expect {
-	    -re "end of file.*$inferior_exited_re normally.*$gdb_prompt $" {
-		pass "send end of file"
+	# Control-D is not EOF on Windows hosts, so skip this test.
+	if { ! [ishost i686-*-mingw32] } {
+	    send_gdb "\004"
+	    gdb_expect {
+		-re "end of file.*$inferior_exited_re normally.*$gdb_prompt $" {
+		    pass "send end of file"
+		}
+		-re "$gdb_prompt $" { fail "send end of file" }
+		timeout { fail "send end of file (timeout)" }
+		eof { fail "send end of file (eof)" }
+	    }
+	} else {	
+	    # Interrupt GDB to get the prompt back.  Killing GDB while the
+	    # remote stub is waiting for file I/O may confuse it for future
+	    # tests.
+	    set msg "Send Control-C to get the prompt"
+	    send_gdb "\003"
+	    gdb_test_multiple "" "$msg" {
+		-re "Program received signal SIGINT.*$gdb_prompt $" {
+		    pass "$msg"
+		}
 	    }
-	    -re "$gdb_prompt $" { fail "send end of file" }
-	    timeout { fail "send end of file (timeout)" }
-	    eof { fail "send end of file (eof)" }
 	}
     }
 }

Re: [PATCH, testsuite] Avoid testsuite harness failure on i686-mingw32 host

[Eli Zaretskii]

> From: Maxim Kuvyrkov <maxim@codesourcery.com>
> Date: Fri, 16 Nov 2012 16:50:50 +1300
> 
> This patch improves testing on i686-mingw32 host.  One of the tests in interrupt.exp assumes that Control-D is EOF, which is not the case for Windows hosts.  After the test timeouts the testsuite doesn't get GDB prompt back, which causes all subsequent tests to fail.

Instead of sending Ctrl-C, isn't it better to send Ctrl-Z, which is
the Windows equivalent of the Posix Ctrl-D?

Re: [PATCH, testsuite] Avoid testsuite harness failure on i686-mingw32 host

[Maxim Kuvyrkov]

On 16/11/2012, at 9:34 PM, Eli Zaretskii wrote:

>> From: Maxim Kuvyrkov <maxim@codesourcery.com>
>> Date: Fri, 16 Nov 2012 16:50:50 +1300
>> 
>> This patch improves testing on i686-mingw32 host.  One of the tests in interrupt.exp assumes that Control-D is EOF, which is not the case for Windows hosts.  After the test timeouts the testsuite doesn't get GDB prompt back, which causes all subsequent tests to fail.
> 
> Instead of sending Ctrl-C, isn't it better to send Ctrl-Z, which is
> the Windows equivalent of the Posix Ctrl-D?

Hm, that's a thought.  I wonder if i686-mingw32-hosted GDB will correctly handle it.

--
Maxim Kuvyrkov
CodeSourcery / Mentor Graphics

Re: [PATCH, testsuite] Avoid testsuite harness failure on i686-mingw32 host

[Kai Tietz]

2012/11/23 Maxim Kuvyrkov <maxim@codesourcery.com>:
> On 16/11/2012, at 9:34 PM, Eli Zaretskii wrote:
>
>>> From: Maxim Kuvyrkov <maxim@codesourcery.com>
>>> Date: Fri, 16 Nov 2012 16:50:50 +1300
>>>
>>> This patch improves testing on i686-mingw32 host.  One of the tests in interrupt.exp assumes that Control-D is EOF, which is not the case for Windows hosts.  After the test timeouts the testsuite doesn't get GDB prompt back, which causes all subsequent tests to fail.
>>
>> Instead of sending Ctrl-C, isn't it better to send Ctrl-Z, which is
>> the Windows equivalent of the Posix Ctrl-D?
>
> Hm, that's a thought.  I wonder if i686-mingw32-hosted GDB will correctly handle it.

Sure, Ctrl-Z (F6) is the proper thing to use.

Kai