group_spawn() on Windows doesn't allow process creation flags to be set

[crwpenn]

When calling group_spawn() like this...

use std::process::Command;
use std::os::windows::process::CommandExt;
use command_group::CommandGroup;

let child = Command::new("app").creation_flags(CREATE_NO_WINDOW).group_spawn();

...I expect the new process to spawn with the CREATE_NO_WINDOW flag applied and to therefore be a background process with no visible window associated. When using spawn(), that's exactly how it happens, but when using group_spawn(), the CREATE_NO_WINDOW flag is ignored and the process spawns with a visible window.

The reason for this behavior is the call to creation_flags() inside group_spawn(), which overrides any process creation flags that were previously set:

command-group/src/stdlib/windows.rs

Line 13 in 9701d77

self.creation_flags(CREATE_SUSPENDED);

If I remove that line from group_spawn(), and then call it like this (to preserve the CREATE_SUSPENDED flag)...

let child = Command::new("app").creation_flags(CREATE_SUSPENDED|CREATE_NO_WINDOW).group_spawn();

...I get the expected result (process spawns with no visible window).

Question: is the CREATE_SUSPENDED flag necessary here? In my use case, at least, it seems not to be - I don't notice any difference between .creation_flags(CREATE_SUSPENDED|CREATE_NO_WINDOW) and .creation_flags(CREATE_NO_WINDOW).

If CREATE_SUSPENDED isn't necessary, could the creation_flags() call be removed from group_spawn() to allow that function to behave the same as spawn() with respect to using creation_flags() to set process creation flags?

If CREATE_SUSPENDED is necessary here, is there a way that it could be added to any previously set flags, instead of overwriting them?

[passcod]

Yeah, it's there to avoid a race condition where the process ends before we assign it to the group.

[passcod]

I'm not really sure what to do, tbh. There's no particularly good way to fetch the flags before writing them. I think the two most viable options are:

• Make a command_group::win::CommandGroupExt trait available only on Windows that provides a group_spawn_with_flags() method, or
• Add an unstable-internals feature that exposes some of the internals of the crate, so you can "make your own" with the flags you want.
  Let me know which you'd prefer.

[crwpenn]

Thanks for the quick response! Given that CREATE_SUSPENDED is needed, I was actually going to suggest the group_spawn_with_flags() solution, so that would be my preference.

I think the best solution would be for std::os::windows::process::CommandExt to have an add_creation_flags() method, so that you could call that instead of creation_flags(), but it looks like that design was considered and rejected.

[passcod]

With the linked issue #17 there's a possibility for a different API which would allow this kind of thing more elegantly

[passcod]

This is now supported (in 2.1.0): https://github.com/watchexec/command-group/blob/main/examples/with_flags.rs

[crwpenn]

Thanks! I like the solution. Craig W. Penner Director Corporate IT Services | The Stream-Flo Group of Companies Office: 780.468.6789 | Direct: 780.468.8539 | Cell:780.902.4009 This email may contain confidential and/or proprietary information. Unauthorized disclosure, copying, or taking action in reliance on the contents is strictly prohibited. Please disregard this email if you are not the original intended recipient. If you received this email in error, please notify the sender. Stream-Flo Industries Ltd. does not accept liability for statements or expressed opinions that are clearly the view of the sender. ♻ Please consider the environment before printing this email.

…

________________________________ From: Félix Saparelli ***@***.***> Sent: Friday, March 3, 2023 4:52:49 AM To: watchexec/command-group ***@***.***> Cc: crwpenn ***@***.***>; Author ***@***.***> Subject: Re: [watchexec/command-group] group_spawn() on Windows doesn't allow process creation flags to be set (Issue #15) Closed #15<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fwatchexec%2Fcommand-group%2Fissues%2F15&data=05%7C01%7C%7C75b49fd5095942d2458108db1bddcfdb%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C638134411709522947%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=W0MvgfQSqjV8zbwvfqwPm0zZL4ybMZqOx0kakoewZDk%3D&reserved=0> as completed. — Reply to this email directly, view it on GitHub<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fwatchexec%2Fcommand-group%2Fissues%2F15%23event-8659066222&data=05%7C01%7C%7C75b49fd5095942d2458108db1bddcfdb%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C638134411709522947%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=L0%2Fr%2FKkxGedIxC0w5b7Miqiy71y6w7gaA5ot2j%2Fx5L4%3D&reserved=0>, or unsubscribe<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FA5YQ24RONYDED3FLLRYFAU3W2HLRDANCNFSM6AAAAAAUXE6JHI&data=05%7C01%7C%7C75b49fd5095942d2458108db1bddcfdb%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C638134411709679199%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=1GI%2BfXn0pC1WQAdxkbG38jrL%2FJahmrLthC7CI5vVdhY%3D&reserved=0>. You are receiving this because you authored the thread.Message ID: ***@***.***>