On Fri, Jun 12, 2020 at 9:00 AM Michael Paquier <michael@paquier.xyz> wrote:
On Tue, Jun 09, 2020 at 11:26:19AM +0100, Dagfinn Ilmari Mannsåker wrote: > Plus a note in the Win32 docs that Win32::Symlink may be required to run > some tests on some Perl/Windows versions..
Planting such a check in individual scripts is not a good idea because it would get forgotten. The best way to handle that is to add a new check in the BEGIN block of TestLib.pm. Note that we already do that with createFile, OsFHandleOpen and CloseHandle. Now the question is: do we really want to make this a hard requirement? I would like to answer yes so as we make sure that this gets always tested, and this needs proper documentation as you say. Now it would be also possible to check if the API is present in the BEGIN block of TestLib.pm, and then use an independent variable similar to what we do with $use_unix_sockets to decide if tests should be skipped or not, but you cannot know if this gets actually, or ever, tested.
The first thing that comes to mind is adding an option to vcregress to choose whether symlinks will be tested or skipped, would that be an acceptable solution?