[1/2] tst_test.sh: Add TST_USES_MODULE

From: Joerg Vehlow <joerg.vehlow@aox-tech.de>

Adds a new library variable TST_USES_MODULE, that can be used, when a
test may need a module, but should not fail, if the module is not available.
---
 doc/test-writing-guidelines.txt |  4 ++-
 testcases/lib/tst_test.sh       | 50 ++++++++++++++++++++++-----------
 2 files changed, 36 insertions(+), 18 deletions(-)

Hi Joerg,

> From: Joerg Vehlow <joerg.vehlow@aox-tech.de>

> Adds a new library variable TST_USES_MODULE, that can be used, when a
> test may need a module, but should not fail, if the module is not available.
I wonder if TST_USES_MODULE is descriptive enough. But it looks to me better
than TST_GET_MODPATH (which Cyril suggested in v3).

We should think twice as _USES_ keyword should be used consistently for the same
approach in different functionality (i.e. TST_USES_FOO is the same as
TST_NEEDS_FOO, but not TCONF/TBROK if it fails).

But whole concept of TST_USES_FOO looks to me a bit complicated, if needed only
for modules. Cannot we just call _tst_find_module directly in this case and not
introduce variable?

...
> +++ b/doc/test-writing-guidelines.txt
> @@ -2125,6 +2125,8 @@ simply by setting right '$TST_NEEDS_FOO'.
>  | 'TST_NEEDS_CMDS'   | String with command names that has to be present for
>                         the test (see below).
>  | 'TST_NEEDS_MODULE' | Test module name needed for the test (see below).
> +| 'TST_USES_MODULE'  | Same as TST_NEEDS_MODULE, except that a missing module
> +|                    | is not an error.
>  | 'TST_NEEDS_DRIVERS'| Checks kernel drivers support for the test.
>  |=============================================================================

> @@ -2174,7 +2176,7 @@ Locating kernel modules
>  +++++++++++++++++++++++

>  The LTP build system can build kernel modules as well, setting
> -'$TST_NEEDS_MODULE' to module name will cause to library to look for the
> +'$TST_NEEDS_MODULE' to module name will cause the library to look for the
This is unrelated change, I merged it as a separate commit (c518ee8b9).

...
> +_tst_find_module()
> +{
> +	local _tst_module=$1
> +	local _tst_is_required=${2:-0}
> +
> +	for tst_module in "$_tst_module" \
> +						"$LTPROOT/testcases/bin/$_tst_module" \
> +						"$TST_STARTWD/$_tst_module"; do
nit: (can be fixed by person who merges it): It's not visible, but uses more
tags than it should be, so it looks like:
+       for tst_module in "$_tst_module" \
+                                               "$LTPROOT/testcases/bin/$_tst_module" \
+                                               "$TST_STARTWD/$_tst_module"; do
+
+                       if [ -f "$tst_module" ]; then
+                               TST_MODPATH="$tst_module"
+                               break
+                       fi
I actually like the original alignment created by Alexey:
        for tst_module in "$TST_NEEDS_MODULE" \
                          "$LTPROOT/testcases/bin/$TST_NEEDS_MODULE" \
                          "$TST_STARTWD/$TST_NEEDS_MODULE"; do

> +
> +			if [ -f "$tst_module" ]; then
> +				TST_MODPATH="$tst_module"
> +				break
> +			fi
> +	done
> +
> +	if [ -z "$TST_MODPATH" ]; then
> +		if [ $_tst_is_required -eq 1 ]; then
> +			tst_brk TCONF "Failed to find module '$_tst_module'"
> +		else
> +			tst_res TINFO "Module '$_tst_module' not found."
nit: please drop dot at the end (can be fixed by person who merges it).
> +		fi
> +	else
> +		tst_res TINFO "Found module at '$TST_MODPATH'"
> +	fi


nit: this is IMHO more readable
	if [ -n "$TST_MODPATH" ]; then
		tst_res TINFO "Found module at '$TST_MODPATH'"
		return
	fi

	if [ $_tst_is_required -eq 1 ]; then
		tst_brk TCONF "Failed to find module '$_tst_module'"
	else
		tst_res TINFO "Module '$_tst_module' not found"
	fi

Kind regards,
Petr

Hi,
>> Adds a new library variable TST_USES_MODULE, that can be used, when a
>> test may need a module, but should not fail, if the module is not available.
> I wonder if TST_USES_MODULE is descriptive enough. But it looks to me better
> than TST_GET_MODPATH (which Cyril suggested in v3).
>
> We should think twice as _USES_ keyword should be used consistently for the same
> approach in different functionality (i.e. TST_USES_FOO is the same as
> TST_NEEDS_FOO, but not TCONF/TBROK if it fails).
>
> But whole concept of TST_USES_FOO looks to me a bit complicated, if needed only
> for modules. Cannot we just call _tst_find_module directly in this case and not
> introduce variable?
I was thinking about adding a function to search for a module, but 
struggled with
returning the name of the found found module.
I had something like
MODPATH=$(tst_find_module "$MODULE_NAME")
but this cannot use tst_res or tst_brk, which I don't like. I also don't 
like just calling
_tst_find_module and getting the result in some "magic" variable. That 
is the reason
why I went with Cyril's Idea of a variable
> ...
>> +_tst_find_module()
>> +{
>> +	local _tst_module=$1
>> +	local _tst_is_required=${2:-0}
>> +
>> +	for tst_module in "$_tst_module" \
>> +						"$LTPROOT/testcases/bin/$_tst_module" \
>> +						"$TST_STARTWD/$_tst_module"; do
> nit: (can be fixed by person who merges it): It's not visible, but uses more
> tags than it should be, so it looks like:
> +       for tst_module in "$_tst_module" \
> +                                               "$LTPROOT/testcases/bin/$_tst_module" \
> +                                               "$TST_STARTWD/$_tst_module"; do
> +
> +                       if [ -f "$tst_module" ]; then
> +                               TST_MODPATH="$tst_module"
> +                               break
> +                       fi
> I actually like the original alignment created by Alexey:
>          for tst_module in "$TST_NEEDS_MODULE" \
>                            "$LTPROOT/testcases/bin/$TST_NEEDS_MODULE" \
>                            "$TST_STARTWD/$TST_NEEDS_MODULE"; do
Just an accident by my editor, I'll fix it for v2
>
>> +
>> +			if [ -f "$tst_module" ]; then
>> +				TST_MODPATH="$tst_module"
>> +				break
>> +			fi
>> +	done
>> +
>> +	if [ -z "$TST_MODPATH" ]; then
>> +		if [ $_tst_is_required -eq 1 ]; then
>> +			tst_brk TCONF "Failed to find module '$_tst_module'"
>> +		else
>> +			tst_res TINFO "Module '$_tst_module' not found."
> nit: please drop dot at the end (can be fixed by person who merges it).
Fixed for v2
>> +		fi
>> +	else
>> +		tst_res TINFO "Found module at '$TST_MODPATH'"
>> +	fi
>
> nit: this is IMHO more readable
> 	if [ -n "$TST_MODPATH" ]; then
> 		tst_res TINFO "Found module at '$TST_MODPATH'"
> 		return
> 	fi
>
> 	if [ $_tst_is_required -eq 1 ]; then
> 		tst_brk TCONF "Failed to find module '$_tst_module'"
> 	else
> 		tst_res TINFO "Module '$_tst_module' not found"
> 	fi
It would still keep the else and not use a return. Indentation clearly 
shows what's going on.
But I agree to invert the logic, first testing the good case, than the bad.
>
> Kind regards,
> Petr

Jörg