On 2022-05-01 Su 09:15, Michael Paquier wrote:
> On Sat, Apr 30, 2022 at 04:52:54PM -0400, Andrew Dunstan wrote:
>> This appears to have been misconceived, at least in the case of MSVC
>> builds, as I have just discovered. It is entirely possible to have the
>> lz4 libraries installed and build with them but not have the .exe, and
>> unlike the configure case the MSVC build system doesn't conduct any test
>> for it, resulting in a nasty looking TAP test failure. I assume similar
>> failures are possible with zstd and maybe gzip.
> Okay, we could change the code of vcregress.pl so as the different ENV
> commands (tar, zstd, lz4 or gzip) are assigned as follows for the sake
> of the tests:
> - If a ENV value is available, trust the environment/user and rely on
> it.
> - If a ENV value is not available, try to look for it in the
> environment by launching a simple $command --version (-version should
> be fine across all the commands currently in need of coverage?).
> -- On failure, set ENV{command} to an empty string.
> -- On success, set ENV{command} = "$command"
>
> Does something like this look better to you?IIRC we know that tar will be available on Windows. I don't think we should do that check for every time we call vc_regress.pl, that seems wasteful. Maybe do it if the command is one that might require these commands, which I think would be bincheck or taptest. And/Or stash some status somewhere? cheers andrew -- Andrew Dunstan EDB: https://www.enterprisedb.com
