[PATCH] Fix test failures when TUI is not enabled

[PATCH] Fix test failures when TUI is not enabled

[Bernd Edlinger]

This adds some missing require allow_tui_tests.

When tui is not enabled those test cases do typically
fail, but most of them only after a long time-out:

FAIL: gdb.base/new-ui.exp: do_test_invalid_args: new-ui with tui
FAIL: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: wrap (timeout)
FAIL: gdb.base/wrap-line.exp: term=ansi: width-auto-detected: wrap (timeout)
FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: $gdb_width == $readline_width
FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: wrap (timeout)
FAIL: gdb.base/wrap-line.exp: term=xterm: width-auto-detected: $gdb_width == $readline_width
FAIL: gdb.base/wrap-line.exp: term=xterm: width-auto-detected: wrap (timeout)
FAIL: gdb.tui/wrap-line.exp: width-hard-coded: cli: wrap
FAIL: gdb.tui/wrap-line.exp: width-hard-coded: cli: prompt after wrap
FAIL: gdb.tui/wrap-line.exp: width-auto-detected: cli: wrap
FAIL: gdb.tui/wrap-line.exp: width-auto-detected: cli: prompt after wrap
---
 gdb/testsuite/gdb.base/new-ui.exp    | 8 +++++---
 gdb/testsuite/gdb.base/wrap-line.exp | 1 +
 gdb/testsuite/gdb.tui/wrap-line.exp  | 1 +
 3 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/gdb/testsuite/gdb.base/new-ui.exp b/gdb/testsuite/gdb.base/new-ui.exp
index 2dfcbf7e108..a56d213e527 100644
--- a/gdb/testsuite/gdb.base/new-ui.exp
+++ b/gdb/testsuite/gdb.base/new-ui.exp
@@ -184,9 +184,11 @@ proc_with_prefix do_test_invalid_args {} {
 	     "new-ui with bad interpreter name"
 
     # Test that the TUI cannot be used for a new UI.
-    gdb_test "new-ui tui $extra_tty_name" \
-	"interpreter 'tui' cannot be used with a new UI" \
-	"new-ui with tui"
+    if [allow_tui_tests] {
+	gdb_test "new-ui tui $extra_tty_name" \
+	    "interpreter 'tui' cannot be used with a new UI" \
+	    "new-ui with tui"
+    }
 
     # Test that we can continue working normally.
     if ![runto_main] {
diff --git a/gdb/testsuite/gdb.base/wrap-line.exp b/gdb/testsuite/gdb.base/wrap-line.exp
index b0931a3bbb8..1a1b47410ee 100644
--- a/gdb/testsuite/gdb.base/wrap-line.exp
+++ b/gdb/testsuite/gdb.base/wrap-line.exp
@@ -18,6 +18,7 @@
 # We set TERM on build, but we need to set it on host.  That only works if
 # build == host.
 require {!is_remote host}
+require allow_tui_tests
 
 # Test both ansi (no auto-wrap) and xterm (auto-wrap).
 set terms {ansi xterm}
diff --git a/gdb/testsuite/gdb.tui/wrap-line.exp b/gdb/testsuite/gdb.tui/wrap-line.exp
index 9cddcf14995..48a04ebc330 100644
--- a/gdb/testsuite/gdb.tui/wrap-line.exp
+++ b/gdb/testsuite/gdb.tui/wrap-line.exp
@@ -17,6 +17,7 @@
 
 # Required for tuiterm.
 require {!is_remote host}
+require allow_tui_tests
 
 tuiterm_env
 
-- 
2.39.2

Re: [PATCH] Fix test failures when TUI is not enabled

[Tom de Vries]

On 8/9/24 10:47, Bernd Edlinger wrote:
> This adds some missing require allow_tui_tests.
> 
> When tui is not enabled those test cases do typically
> fail, but most of them only after a long time-out:
> 
> FAIL: gdb.base/new-ui.exp: do_test_invalid_args: new-ui with tui
> FAIL: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: wrap (timeout)
> FAIL: gdb.base/wrap-line.exp: term=ansi: width-auto-detected: wrap (timeout)
> FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: $gdb_width == $readline_width
> FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: wrap (timeout)
> FAIL: gdb.base/wrap-line.exp: term=xterm: width-auto-detected: $gdb_width == $readline_width
> FAIL: gdb.base/wrap-line.exp: term=xterm: width-auto-detected: wrap (timeout)
> FAIL: gdb.tui/wrap-line.exp: width-hard-coded: cli: wrap
> FAIL: gdb.tui/wrap-line.exp: width-hard-coded: cli: prompt after wrap
> FAIL: gdb.tui/wrap-line.exp: width-auto-detected: cli: wrap
> FAIL: gdb.tui/wrap-line.exp: width-auto-detected: cli: prompt after wrap
> ---

Hi,

I disabled tui support using --disable-tui, and can't reproduce the 
wrap-line fails you see.

Is it possible that you're running into "[gdb/testsuite, stub-termcap] 
FAIL: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: wrap 
(timeout)" @ https://sourceware.org/bugzilla/show_bug.cgi?id=31201 ?

Anyway, adding require allow_tui_tests in those two test-cases looks wrong.

In the case of gdb.tui/wrap-line.exp, there is already a test that bails
out if TUI is unsupported, but still allows testing of things that 
should work regardless.

In the case of gdb.base/wrap-line.exp, it's a test-case I added 
specifically to test the behaviour of prompt wrapping in the CLI, not in 
TUI, so requiring allow_tui_tests there is non-sensical.

The change in gdb.base/new-ui.exp LGTM.  You could submit a v2 with just 
that change.

Thanks,
- Tom

>   gdb/testsuite/gdb.base/new-ui.exp    | 8 +++++---
>   gdb/testsuite/gdb.base/wrap-line.exp | 1 +
>   gdb/testsuite/gdb.tui/wrap-line.exp  | 1 +
>   3 files changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/gdb/testsuite/gdb.base/new-ui.exp b/gdb/testsuite/gdb.base/new-ui.exp
> index 2dfcbf7e108..a56d213e527 100644
> --- a/gdb/testsuite/gdb.base/new-ui.exp
> +++ b/gdb/testsuite/gdb.base/new-ui.exp
> @@ -184,9 +184,11 @@ proc_with_prefix do_test_invalid_args {} {
>   	     "new-ui with bad interpreter name"
>   
>       # Test that the TUI cannot be used for a new UI.
> -    gdb_test "new-ui tui $extra_tty_name" \
> -	"interpreter 'tui' cannot be used with a new UI" \
> -	"new-ui with tui"
> +    if [allow_tui_tests] {
> +	gdb_test "new-ui tui $extra_tty_name" \
> +	    "interpreter 'tui' cannot be used with a new UI" \
> +	    "new-ui with tui"
> +    }
>   
>       # Test that we can continue working normally.
>       if ![runto_main] {
> diff --git a/gdb/testsuite/gdb.base/wrap-line.exp b/gdb/testsuite/gdb.base/wrap-line.exp
> index b0931a3bbb8..1a1b47410ee 100644
> --- a/gdb/testsuite/gdb.base/wrap-line.exp
> +++ b/gdb/testsuite/gdb.base/wrap-line.exp
> @@ -18,6 +18,7 @@
>   # We set TERM on build, but we need to set it on host.  That only works if
>   # build == host.
>   require {!is_remote host}
> +require allow_tui_tests
>   
>   # Test both ansi (no auto-wrap) and xterm (auto-wrap).
>   set terms {ansi xterm}
> diff --git a/gdb/testsuite/gdb.tui/wrap-line.exp b/gdb/testsuite/gdb.tui/wrap-line.exp
> index 9cddcf14995..48a04ebc330 100644
> --- a/gdb/testsuite/gdb.tui/wrap-line.exp
> +++ b/gdb/testsuite/gdb.tui/wrap-line.exp
> @@ -17,6 +17,7 @@
>   
>   # Required for tuiterm.
>   require {!is_remote host}
> +require allow_tui_tests
>   
>   tuiterm_env
>   

Re: [PATCH] Fix test failures when TUI is not enabled

[Bernd Edlinger]

On 8/9/24 11:28, Tom de Vries wrote:
> On 8/9/24 10:47, Bernd Edlinger wrote:
>> This adds some missing require allow_tui_tests.
>>
>> When tui is not enabled those test cases do typically
>> fail, but most of them only after a long time-out:
>>
>> FAIL: gdb.base/new-ui.exp: do_test_invalid_args: new-ui with tui
>> FAIL: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: wrap (timeout)
>> FAIL: gdb.base/wrap-line.exp: term=ansi: width-auto-detected: wrap (timeout)
>> FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: $gdb_width == $readline_width
>> FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: wrap (timeout)
>> FAIL: gdb.base/wrap-line.exp: term=xterm: width-auto-detected: $gdb_width == $readline_width
>> FAIL: gdb.base/wrap-line.exp: term=xterm: width-auto-detected: wrap (timeout)
>> FAIL: gdb.tui/wrap-line.exp: width-hard-coded: cli: wrap
>> FAIL: gdb.tui/wrap-line.exp: width-hard-coded: cli: prompt after wrap
>> FAIL: gdb.tui/wrap-line.exp: width-auto-detected: cli: wrap
>> FAIL: gdb.tui/wrap-line.exp: width-auto-detected: cli: prompt after wrap
>> ---
> 
> Hi,
> 
> I disabled tui support using --disable-tui, and can't reproduce the wrap-line fails you see.
> 
> Is it possible that you're running into "[gdb/testsuite, stub-termcap] FAIL: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: wrap (timeout)" @ https://sourceware.org/bugzilla/show_bug.cgi?id=31201 ?
> 

Yes, that could be the case.

I did not explicitly opt-out tui, I just did never care to install
e.g. libcurses etc. and the configure did the rest.

> Anyway, adding require allow_tui_tests in those two test-cases looks wrong.
> 

Okay, but what is going on then?

> In the case of gdb.tui/wrap-line.exp, there is already a test that bails
> out if TUI is unsupported, but still allows testing of things that should work regardless.
> 
> In the case of gdb.base/wrap-line.exp, it's a test-case I added specifically to test the behaviour of prompt wrapping in the CLI, not in TUI, so requiring allow_tui_tests there is non-sensical.
> 

Well is there a better way to detect the issue without going into this long timeout,
or can it be fixed somehow?

> The change in gdb.base/new-ui.exp LGTM.  You could submit a v2 with just that change.
> 

OK, will do.

Thanks
Bernd.

Re: [PATCH] Fix test failures when TUI is not enabled

[Bernd Edlinger]

On 8/9/24 11:48, Bernd Edlinger wrote:
> On 8/9/24 11:28, Tom de Vries wrote:
>> On 8/9/24 10:47, Bernd Edlinger wrote:
>>
>> I disabled tui support using --disable-tui, and can't reproduce the wrap-line fails you see.
>>
>> Is it possible that you're running into "[gdb/testsuite, stub-termcap] FAIL: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: wrap (timeout)" @ https://sourceware.org/bugzilla/show_bug.cgi?id=31201 ?
>>
> 
> Yes, that could be the case.
> 
> I did not explicitly opt-out tui, I just did never care to install
> e.g. libcurses etc. and the configure did the rest.
> 

FYI this is what happens in the test:

PASS: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: $gdb_width == [expr $readline_width + 1]
PASS: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: width
print 1^M
$1 = 1^M
(gdb) 78901234567890123456789012345678901234567890123456789012345678901234567^M<890123456789012345678901234567890123456789012345678                         ^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H9WFAIL: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: wrap (timeout)
Quit^M


the input is long string and the output is
^M<890123456789012345678901234567890123456789012345678
+ some space characters + some backspace characters
and no (gdb) prompt.  That is not what the test expects?


Thanks
Bernd.

Re: [PATCH] Fix test failures when TUI is not enabled

[Tom de Vries]

On 8/9/24 11:48, Bernd Edlinger wrote:
> On 8/9/24 11:28, Tom de Vries wrote:
>> On 8/9/24 10:47, Bernd Edlinger wrote:
>>> This adds some missing require allow_tui_tests.
>>>
>>> When tui is not enabled those test cases do typically
>>> fail, but most of them only after a long time-out:
>>>
>>> FAIL: gdb.base/new-ui.exp: do_test_invalid_args: new-ui with tui
>>> FAIL: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: wrap (timeout)
>>> FAIL: gdb.base/wrap-line.exp: term=ansi: width-auto-detected: wrap (timeout)
>>> FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: $gdb_width == $readline_width
>>> FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: wrap (timeout)
>>> FAIL: gdb.base/wrap-line.exp: term=xterm: width-auto-detected: $gdb_width == $readline_width
>>> FAIL: gdb.base/wrap-line.exp: term=xterm: width-auto-detected: wrap (timeout)
>>> FAIL: gdb.tui/wrap-line.exp: width-hard-coded: cli: wrap
>>> FAIL: gdb.tui/wrap-line.exp: width-hard-coded: cli: prompt after wrap
>>> FAIL: gdb.tui/wrap-line.exp: width-auto-detected: cli: wrap
>>> FAIL: gdb.tui/wrap-line.exp: width-auto-detected: cli: prompt after wrap
>>> ---
>>
>> Hi,
>>
>> I disabled tui support using --disable-tui, and can't reproduce the wrap-line fails you see.
>>
>> Is it possible that you're running into "[gdb/testsuite, stub-termcap] FAIL: gdb.base/wrap-line.exp: term=ansi: width-hard-coded: wrap (timeout)" @ https://sourceware.org/bugzilla/show_bug.cgi?id=31201 ?
>>
> 
> Yes, that could be the case.
> 
> I did not explicitly opt-out tui, I just did never care to install
> e.g. libcurses etc. and the configure did the rest.

That sounds like you ran into that PR then.

>> Anyway, adding require allow_tui_tests in those two test-cases looks wrong.
>>
> 
> Okay, but what is going on then?
> 

We can get the same effect using "TERM=dumb gdb".

It forces readline's horizontal-scroll-mode to activate (unfortunately, 
without affecting the value of the corresponding variable that we can 
read with bind -v).

>> In the case of gdb.tui/wrap-line.exp, there is already a test that bails
>> out if TUI is unsupported, but still allows testing of things that should work regardless.
>>
>> In the case of gdb.base/wrap-line.exp, it's a test-case I added specifically to test the behaviour of prompt wrapping in the CLI, not in TUI, so requiring allow_tui_tests there is non-sensical.
>>
> 
> Well is there a better way to detect the issue without going into this long timeout,
> or can it be fixed somehow?
> 

We could accept the string produced by horizontal-scroll-mode and 
produce an unsupported and bail out.

Thanks,
- Tom

>> The change in gdb.base/new-ui.exp LGTM.  You could submit a v2 with just that change.
>>
> 
> OK, will do.
> 
> Thanks
> Bernd.
> 

Re: [PATCH] Fix test failures when TUI is not enabled

[Bernd Edlinger]

On 8/9/24 13:17, Tom de Vries wrote:
> 
> That sounds like you ran into that PR then.
> 
>>> Anyway, adding require allow_tui_tests in those two test-cases looks wrong.
>>>
>>
>> Okay, but what is going on then?
>>
> 
> We can get the same effect using "TERM=dumb gdb".
> 
> It forces readline's horizontal-scroll-mode to activate (unfortunately, without affecting the value of the corresponding variable that we can read with bind -v).
> 
>>> In the case of gdb.tui/wrap-line.exp, there is already a test that bails
>>> out if TUI is unsupported, but still allows testing of things that should work regardless.
>>>
>>> In the case of gdb.base/wrap-line.exp, it's a test-case I added specifically to test the behaviour of prompt wrapping in the CLI, not in TUI, so requiring allow_tui_tests there is non-sensical.
>>>
>>
>> Well is there a better way to detect the issue without going into this long timeout,
>> or can it be fixed somehow?
>>
> 
> We could accept the string produced by horizontal-scroll-mode and produce an unsupported and bail out.
> 


Hmm...,


--- a/gdb/testsuite/gdb.base/wrap-line.exp
+++ b/gdb/testsuite/gdb.base/wrap-line.exp
@@ -115,6 +115,9 @@ proc test_wrap { width_auto_detected } {
        -re $re {
            pass $gdb_test_name
        }
+       -re "\r<.*" {
+           unsupported $gdb_test_name
+       }
     }
 
     # Generate a prompt.


This works as expected, but I see no way how to do the same
in gdb.tui/wrap-line.exp because the 

... and other than that there is also
FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: $gdb_width == $readline_width
FAIL: gdb.base/wrap-line.exp: term=xterm: width-auto-detected: $gdb_width == $readline_width

(gdb) PASS: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: set width 80
show editing^M
Editing of command lines as they are typed is on.^M
(gdb) maint info screen^M
Number of characters gdb thinks are in a line is 80.^M
Number of characters readline reports are in a line is 79.^M
Number of characters environment thinks are in a line is 80 (COLUMNS).^M
Number of lines gdb thinks are in a page is 4294967295 (unlimited).^M
Number of lines readline reports are in a page is 32767 (unlimited).^M
Number of lines environment thinks are in a page is 25 (LINES).^M
(gdb) PASS: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: maint info screen
FAIL: gdb.base/wrap-line.exp: term=xterm: width-hard-coded: $gdb_width == $readline_width

I wonder whether that can be used to detect this situation,
when bind -v is not able to it?


Thanks
Bernd.

Re: [PATCH] Fix test failures when tui is not enabled

[Bernd Edlinger]

Okay,

although I did not expect to get it solved, it looks like a possible fix.

please have a look at my patch here:
https://sourceware.org/pipermail/gdb-patches/2024-August/210963.html

Thanks
Bernd.