Re: [PATCH v2] app/testpmd: show output of commands read from file

[Ferruh Yigit <ferruh.yigit@amd.com>]

On 8/22/2024 6:18 PM, Bruce Richardson wrote:
> On Thu, Aug 22, 2024 at 06:14:55PM +0100, Bruce Richardson wrote:
>> On Thu, Aug 22, 2024 at 05:53:27PM +0100, Ferruh Yigit wrote:
>>> On 8/22/2024 11:41 AM, Bruce Richardson wrote:
>>>> Testpmd supports the "--cmdline-file" parameter to read a set of initial
>>>> commands from a file. However, the only indication that this has been
>>>> done successfully on startup is a single-line message, no output from
>>>> the commands is seen.
>>>>
>>>
>>> For user I think it makes sense to see the command [1], only concern is
>>> if someone parsing testpmd output may be impacted on this, although I
>>> expect it should be trivial to update the relevant parsing.
>>>
>>> [1]
>>> Btw, I can still see the command output, I assume because command does
>>> the printf itself, for example for 'show port summary 0' command:
>>> - before patch:
>>> ...
>>> Number of available ports: 2
>>> Port MAC Address       Name         Driver         Status   Link
>>> 0    xx:xx:xx:xx:xx:xx xxxx:xx:xx.x aaaaaaaa       up       xxx Gbps
>>> ...
>>>
>>> - after patch
>>> ...
>>> testpmd> show port summary 0
>>> Number of available ports: 2
>>> Port MAC Address       Name         Driver         Status   Link
>>> 0    xx:xx:xx:xx:xx:xx xxxx:xx:xx.x aaaaaaaa       up       xxx Gbps
>>> ...
>>>
>>> Only difference above is, after patch the command itself also printed.
>>>
>>>
>>
>> That's because the function uses printf itself, which is actually wrong.
>> Any output from a cmdline function should use the "cmdline_printf" call
>> which outputs to the proper cmdline filehandle.
>>

Got it.
But in existing testpmd code, only a handful cmdline functions use the
'cmdline_printf' and most of them are in the same help function.
At this stage I think no need to update them. There is already some
confusion on testpmd logging between printf & TESTPMD_LOG().

>>>> To improve usability here, we can use cmdline_new rather than
>>>> cmdline_file_new and have the output from the various commands sent to
>>>> stdout, allowing the user to see better what is happening.
>>>>
>>>> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
>>>>
>>>> ---
>>>> v2: use STDOUT_FILENO in place of hard-coded "1"
>>>> ---
>>>>  app/test-pmd/cmdline.c | 14 +++++++++++++-
>>>>  1 file changed, 13 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
>>>> index b7759e38a8..52e64430d9 100644
>>>> --- a/app/test-pmd/cmdline.c
>>>> +++ b/app/test-pmd/cmdline.c
>>>> @@ -6,6 +6,7 @@
>>>>  #include <ctype.h>
>>>>  #include <stdarg.h>
>>>>  #include <errno.h>
>>>> +#include <fcntl.h>
>>>>  #include <stdio.h>
>>>>  #include <stdint.h>
>>>>  #include <stdlib.h>
>>>> @@ -13431,7 +13432,18 @@ cmdline_read_from_file(const char *filename)
>>>>  {
>>>>  	struct cmdline *cl;
>>>>  
>>>> -	cl = cmdline_file_new(main_ctx, "testpmd> ", filename);
>>>> +	/* cmdline_file_new does not produce any output which is not ideal here.
>>>> +	 * Much better to show output of the commands, so we open filename directly
>>>> +	 * and then pass that to cmdline_new with stdout as the output path.
>>>> +	 */
>>>> +	int fd = open(filename, O_RDONLY);
>>>> +	if (fd < 0) {
>>>> +		fprintf(stderr, "Failed to open file %s: %s\n",
>>>> +			filename, strerror(errno));
>>>> +		return;
>>>> +	}
>>>> +
>>>> +	cl = cmdline_new(main_ctx, "testpmd> ", fd, STDOUT_FILENO);
>>>>
>>>
>>> Above is almost save as 'cmdline_file_new()' function with only
>>> difference that it uses '-1' for 's_out'.
>>>
>>> If this usecase may be required by others, do you think does it have a
>>> value to pass 's_out' to 'cmdline_file_new()' or have a new version of
>>> API that accepts 's_out' as parameter?
>>>
>>
>> Yes, I thought about this, and actually started implementing a new API for
>> cmdline library to that. However, I decided that, given the complexity
>> here, that it's not really necessary - especially as there is no clear way
>> to do things. The options are:
>>
>> * extend cmdline_file_new to have a flag to echo to stdout (which would be
>>   the very common case here).
>> * extend cmdline_file_new to take a file handle - this is more flexible,
>>   but slightly less usable.
>> * add a new cmdline_file_<something>_new function to echo to stdout.
>> * add a new cmdline_file_<something>_new function to write to a filehandle.
>>
>> I don't like breaking the cmdline API (and ABI), so I didn't want to do
>> either #1 or #2, which would be the cleanest solutions. For #3 and #4,
>> naming is hard, and deciding between them is even harder. Given the choice,
>> I prefer #3, as I can't see #4 being very common and we always have
>> cmdline_new as a fallback anyway.
>>
>> Overall, though, I threw away that work, because it didn't seem worth it,
>> for the sake of having the user to do an extra "open" call.
>>
> 

I vote to option #1, but agree that it may not worth breaking API and ABI.

Is 'cmdline_file_new_v2()' too bad a name, perhaps better to go with
testpmd implementation, as you did in this patch.

> And also to add:
> If there is clear consensus on what the correct option for this case is,
> I'm happy enough to go back and extend the cmdline library as agreed.
> :-)