PHPT ===DONE===

Gregory Beaver schrieb:

Hello Gregory,

after a while, spending time on this code, I'm back...


> > Are there any infos on the PEAR website about phpt? For example
> > RunTest.php knows the sections "FAIL" and "RETURNS", which aren't
> > described on [1].

"UPLOAD" is also a section not really used in "run-tests.php". But a good
idea.


> --RETURNS-- is used to test the return value for scripts that are used
> in shell scripts.

That's also a good idea. But why want you have this feature only on Windows?


There are also some other problems. Let me remmeber:
- ENV and COOKIE and test whitch require the cgi-binary are not working with
PHP4
- RETURNS does not work with all PHP binarys (depends on the configure
options) and OS
- POST_RAW does not allow a real raw http message, according to the rfc
- does not work correct on Win9x
- does not work with spaces in pathnames
- does not work if the phpt files uses "\r" for lineendings
- does not work if you use the cgi binary for pear.
- the second parameter from run() can only be used for the PhpUnit test
- tap output is not correct
- wrong test summary if you use top output
(I hope, I have not forgotten one...)

So in the current form, I don't think you can recommend this test [1].


Another question:
Why is the value for the doc block tag "@category" "pear" (in all PEAR
files)? If I read this [2] correct, it should be "PEAR", or not?

Regards,
Carsten

[1]
http://pear.php.net/manual/en/develo...g.requirements
[2] http://pear.php.net/manual/en/develo...ting.howto.php

Carsten Wiedmann wrote:
>> --RETURNS-- is used to test the return value for scripts that are used
>> in shell scripts.
>
> That's also a good idea. But why want you have this feature only on
> Windows?

???? It works on all OSes.

> There are also some other problems. Let me remmeber:
> - ENV and COOKIE and test whitch require the cgi-binary are not working
> with PHP4
> - RETURNS does not work with all PHP binarys (depends on the configure
> options) and OS
> - POST_RAW does not allow a real raw http message, according to the rfc
> - does not work correct on Win9x

Win9x is not supported by PHP and I do not have a working Win9x box, the
last one died over 3 years ago.

> - does not work with spaces in pathnames
> - does not work if the phpt files uses "\r" for lineendings
> - does not work if you use the cgi binary for pear.
> - the second parameter from run() can only be used for the PhpUnit test
> - tap output is not correct
> - wrong test summary if you use top output
> (I hope, I have not forgotten one...)

If these are true, then it is your responsibility to open bugs. This is
open source and I cannot use my considerable powers of ESP to divine al
possible bugs, only most of them .

Seriuosly though, I have had absolutely no trouble running unit tests on
windows, OS X, freebsd and linux. I personally use the --cgi option on
windows and linux regularly. Additionally, the code is a direct port of
the code in php-src/runtests.php - don't you think that someone would
have noticed before if the code was broken for any other test
environments? Perhaps your system is unique.

TAP output is valid, I don't know what you're talking about.

Frankly your list is frustrating because none of the claims are
substantiated with a test, and there isn't a single open bug for any of
these issues. Please put the proof out for all to see and I will take
these claims more seriously.

> So in the current form, I don't think you can recommend this test [1].

Au contraire. It is an extremely battle-tested, robust testing format
responsible for oodles of great testing over the past several years.

> Another question:
> Why is the value for the doc block tag "@category" "pear" (in all PEAR
> files)? If I read this [2] correct, it should be "PEAR", or not?

phpDocumentor lower-cases this value in the only converter that needs
it, the XMLocBook/peardoc2 converter

Greg

Greg Beaver schrieb:

> > > --RETURNS-- is used to test the return value for scripts that are used
> > > in shell scripts.
> >
> > That's also a good idea. But why want you have this feature only on
> > Windows?
>
> ???? It works on all OSes.

No, it's just the way you have provided the expected output too. Have you
ever changed only the return value, for example in
".\tests\PEAR_Command_Registry\shell-test\test_rel_ne.phpt", from "0" to "1"
on a Unix system?

(BTW: On Unix, PHP may not be built whith --enable-sigchild if you want use
RETURNS with the current code)


> Win9x is not supported by PHP and I do not have a working Win9x box, the
> last one died over 3 years ago.

That's wrong [1]. But the PEAR command line tool does not work on Win9x.
That's true.

BTW:
I know a lot of people, which allways (must) use Win9x. And they are also
interested in learning/using PHP (and maybe PEAR).


> If these are true, then it is your responsibility to open bugs. This is
> open source and I cannot use my considerable powers of ESP to divine al
> possible bugs, only most of them .

Sorry, I thought you know about these problems. And so this is at first a
list for information.
OK, I can open bug reports, or better one. Many things depends on each
other.

For me I have now a working version of this tool. But it's a little bit away
from your code... I can upload my current version (with a diff) to a
webserver, so you or someone other can look on this.


> Seriuosly though, I have had absolutely no trouble running unit tests on
> windows, OS X, freebsd and linux.

But I had these problems. And that was the point I have start my work with
this code.


> I personally use the --cgi option on windows and linux regularly.

But not with PHP4. You are using PHP5, or not?
--> The current test can't use the environment with PHP4.


> Additionally, the code is a direct port of
> the code in php-src/runtests.php -

IMHO that's the problem. They are using this test to test a specific PHP
version itself. And I (or we) want use it, to test if our php code is
working correct (under several conditions (os/php versions)).


> TAP output is valid, I don't know what you're talking about.

| node2# bin/pear run-tests -t PEAR_RunTest/
| Running 3 tests
| FAIL PEAR_RunTest --INI--[filename]
| SKIP PEAR_RunTest --POST--[filename] (reason: ...)
| PASS PEAR_RunTest improve diff of failing EXPECTF test[filename]
| wrote TAP-format log to "run-tests.log"
| TOTAL TIME: 00:07
| 0 PASSED TESTS
| 0 SKIPPED TESTS
--^ <-- here

| node2# cat run-tests.log
| 1..3
| # Expected output:
| #
| #tring(5) "hooba"#
| #
| # Actual output:
| #
| #string(5) "hooba"not ok 1 - PEAR_RunTest --INI--[filename]
--------------------^ <--- here

| ok 2 # skip --cgi ...
| ok 3 - PEAR_RunTest improve diff of failing EXPECTF test[filename]
| node2#

I don't think that this is correct.


> > - does not work with spaces in pathnames
> > - does not work with spaces in pathnames
> > - does not work if the phpt files uses "\r" for lineendings
> > - does not work if you use the cgi binary for pear.
> > - the second parameter from run() can only be used for the PhpUnit test
> Please put the proof out for all to see and I will take
> these claims more seriously.

Better you try this points yourself... It's hard to show this. Or want you
have a copy 'n paste from the shell for each point?


> > So in the current form, I don't think you can recommend this test [1].
>
> Au contraire. It is an extremely battle-tested, robust testing format
> responsible for oodles of great testing over the past several years.

But it's a nice tool and easy to use. The only problem was, it have not
worked on all /my/ test installations/environments.


> > Another question:
> > Why is the value for the doc block tag "@category" "pear" (in all PEAR
> > files)? If I read this [2] correct, it should be "PEAR", or not?
>
> phpDocumentor lower-cases this value in the only converter that needs
> it, the XMLocBook/peardoc2 converter

Thanks, so I have to change this back in my code. (phpcs also mark this as
an error)


Regards,
Carsten

[1] http://de.php.net/manual/en/install.windows.php