RE: [PHP-CVS] com php-src: Add VT100 support for Windows: ext/standard/basic_functions.c ext/standard/streamsfuncs.c ext/standard/streamsfuncs.h
run-tests.php sapi/cli/php_cli.c tests/output/sapi_windows_vt100_support.inc tests/output/sapi_windows_vt100_s

Hi Matteo, > -----Original Message----- > From: Matteo Beccati [mailto:php@beccati.com] > Sent: Thursday, November 3, 2016 2:37 PM > To: Anatol Belski <anatol.php@belski.net> > Cc: php-cvs@lists.php.net; 'Michele Locati' <michele@locati.it> > Subject: Re: [PHP-CVS] com php-src: Add VT100 support for Windows: > ext/standard/basic_functions.c ext/standard/streamsfuncs.c > ext/standard/streamsfuncs.h run-tests.php sapi/cli/php_cli.c > tests/output/sapi_windows_vt100_support.inc > tests/output/sapi_windows_vt100_s > > Hi Anatol, Michele, > > On 31/10/2016 21:41, Anatol Belski wrote: > > Matteo, seems I've no choice than asking you to do some check in this > > regard directly on the CI box, as it's hard to move forward without a > > reproducer. Or maybe Michele has a better idea, while looking at the > > fails. > > I also am not able to replicate it via command line, as it seems Bamboo does > things differently when running the command. However, running the test via ssh > produces the same result: > > ssh target.host.name php path/to/run-tests.php -p /path/to/sapi/cli/php > /path/to/tests/output/stream_isatty_out.phpt --show-diff > That's the catch. Of course, if there's yet one IO redirection level above the top level PHP process, it'll hinder the actual test. I can now reproduce it, no SSH is needed. Based on yours echo php path/to/run-tests.php -p /path/to/sapi/cli/php /path/to/tests/output/stream_isatty_out.phpt --show-diff | php -r '$s = fread(STDIN, 2048); system($s);' This makes the test to behave wrong, whereby the result of stream_isatty is correct as the invoker of run-tests.php is not the top level process anymore. So this is a false positive caused by a specific way the test is run. I think, it might make sense to introduce another skipif group, like SKIP_IO_REDIRECTION_TESTS. Until one could affect the way a particular CI runs tests, seems it's the only option to avoid false positive reports on this kind of test. Regards Anatol