Fix cursor remains busy after launching windowed app

[Jesse205]

This PR fixes a regression where the cursor remains stuck in a busy state.

The _WINDOWS macro is not defined in the pymanager project; it should be PY_WINDOWED. Therefore, I replaced defined(_WINDOWS) with PY_WINDOWED.

This resolves a previously fixed issue that has resurfaced: python/cpython#61492

---

Update: Relocated code to maintain a continuous cursor busy state during the pymanager-to-interpreter startup sequence.

[zooba]

Good catch. I think we also need to define the variable for the other kind of launcher, too, which lives in _msbuild.py in the top level directory. Want to take a look at that?

[Jesse205]

Sorry, I'm not sure what you're referring to. In pymanager, _WINDOWS only appears here (this is the only location I could find via search).

[zooba]

The _msbuild.py file in the top of the repo is our build script, and it's where Py_WINDOWED is defined. But it's not been defined for the launcherw.exe (template that gets renamed for generating shortcuts), which means these ones won't benefit from this fix.

I can get to it when I get a chance, but it'd be good to fix it all in one PR.

[Jesse205]

launcherw.exe appears to work correctly even without PY_WINDOWED. I just installed Python 3.14 using the released pymanager v26.1 and dragged my .pyzw file onto pythonw3.14.exe. The cursor behaves as expected. Using Process Explorer, I examined the process tree and found that there are no child processes under this pythonw3.14.exe. Additionally, the window it creates is exactly the one generated by my .pyzw file. This seems to differ from how pymanager works.

[zooba]

For runtimes with a compatible python3.dll, yes, we bypass the child process for speed (and can't do it for py.exe because that's already loaded one Python runtime). But if python3.dll is missing, or too old, then it'll launch the executable and we're in the same state.

[zooba]

I'll just take this as it is and file a new issue to fix the other launcher. Thanks for the contribution!

[Jesse205]

This PR fixes a regression where the cursor remains stuck in a busy state.

The _WINDOWS macro is not defined in the pymanager project; it should be PY_WINDOWED. Therefore, I replaced defined(_WINDOWS) with PY_WINDOWED.

This resolves a previously fixed issue that has resurfaced: python/cpython#61492

---

Update: Relocated code to maintain a continuous cursor busy state during the pymanager-to-interpreter startup sequence.

[zooba]

Good catch. I think we also need to define the variable for the other kind of launcher, too, which lives in _msbuild.py in the top level directory. Want to take a look at that?

[Jesse205]

Sorry, I'm not sure what you're referring to. In pymanager, _WINDOWS only appears here (this is the only location I could find via search).

[zooba]

The _msbuild.py file in the top of the repo is our build script, and it's where Py_WINDOWED is defined. But it's not been defined for the launcherw.exe (template that gets renamed for generating shortcuts), which means these ones won't benefit from this fix.

I can get to it when I get a chance, but it'd be good to fix it all in one PR.

[Jesse205]

launcherw.exe appears to work correctly even without PY_WINDOWED. I just installed Python 3.14 using the released pymanager v26.1 and dragged my .pyzw file onto pythonw3.14.exe. The cursor behaves as expected. Using Process Explorer, I examined the process tree and found that there are no child processes under this pythonw3.14.exe. Additionally, the window it creates is exactly the one generated by my .pyzw file. This seems to differ from how pymanager works.

[zooba]

For runtimes with a compatible python3.dll, yes, we bypass the child process for speed (and can't do it for py.exe because that's already loaded one Python runtime). But if python3.dll is missing, or too old, then it'll launch the executable and we're in the same state.

[zooba]

I'll just take this as it is and file a new issue to fix the other launcher. Thanks for the contribution!

[Jesse205]

This PR fixes a regression where the cursor remains stuck in a busy state.

The _WINDOWS macro is not defined in the pymanager project; it should be PY_WINDOWED. Therefore, I replaced defined(_WINDOWS) with PY_WINDOWED.

This resolves a previously fixed issue that has resurfaced: python/cpython#61492

---

Update: Relocated code to maintain a continuous cursor busy state during the pymanager-to-interpreter startup sequence.

[zooba]

Good catch. I think we also need to define the variable for the other kind of launcher, too, which lives in _msbuild.py in the top level directory. Want to take a look at that?

[Jesse205]

Sorry, I'm not sure what you're referring to. In pymanager, _WINDOWS only appears here (this is the only location I could find via search).

[zooba]

The _msbuild.py file in the top of the repo is our build script, and it's where Py_WINDOWED is defined. But it's not been defined for the launcherw.exe (template that gets renamed for generating shortcuts), which means these ones won't benefit from this fix.

I can get to it when I get a chance, but it'd be good to fix it all in one PR.

[Jesse205]

launcherw.exe appears to work correctly even without PY_WINDOWED. I just installed Python 3.14 using the released pymanager v26.1 and dragged my .pyzw file onto pythonw3.14.exe. The cursor behaves as expected. Using Process Explorer, I examined the process tree and found that there are no child processes under this pythonw3.14.exe. Additionally, the window it creates is exactly the one generated by my .pyzw file. This seems to differ from how pymanager works.

[zooba]

For runtimes with a compatible python3.dll, yes, we bypass the child process for speed (and can't do it for py.exe because that's already loaded one Python runtime). But if python3.dll is missing, or too old, then it'll launch the executable and we're in the same state.

[zooba]

I'll just take this as it is and file a new issue to fix the other launcher. Thanks for the contribution!

[Jesse205]

This PR fixes a regression where the cursor remains stuck in a busy state.

The _WINDOWS macro is not defined in the pymanager project; it should be PY_WINDOWED. Therefore, I replaced defined(_WINDOWS) with PY_WINDOWED.

This resolves a previously fixed issue that has resurfaced: python/cpython#61492

---

Update: Relocated code to maintain a continuous cursor busy state during the pymanager-to-interpreter startup sequence.

[zooba]

Good catch. I think we also need to define the variable for the other kind of launcher, too, which lives in _msbuild.py in the top level directory. Want to take a look at that?

[Jesse205]

Sorry, I'm not sure what you're referring to. In pymanager, _WINDOWS only appears here (this is the only location I could find via search).

[zooba]

The _msbuild.py file in the top of the repo is our build script, and it's where Py_WINDOWED is defined. But it's not been defined for the launcherw.exe (template that gets renamed for generating shortcuts), which means these ones won't benefit from this fix.

I can get to it when I get a chance, but it'd be good to fix it all in one PR.

[Jesse205]

launcherw.exe appears to work correctly even without PY_WINDOWED. I just installed Python 3.14 using the released pymanager v26.1 and dragged my .pyzw file onto pythonw3.14.exe. The cursor behaves as expected. Using Process Explorer, I examined the process tree and found that there are no child processes under this pythonw3.14.exe. Additionally, the window it creates is exactly the one generated by my .pyzw file. This seems to differ from how pymanager works.

[zooba]

For runtimes with a compatible python3.dll, yes, we bypass the child process for speed (and can't do it for py.exe because that's already loaded one Python runtime). But if python3.dll is missing, or too old, then it'll launch the executable and we're in the same state.

[zooba]

I'll just take this as it is and file a new issue to fix the other launcher. Thanks for the contribution!

[Jesse205]

This PR fixes a regression where the cursor remains stuck in a busy state.

The _WINDOWS macro is not defined in the pymanager project; it should be PY_WINDOWED. Therefore, I replaced defined(_WINDOWS) with PY_WINDOWED.

This resolves a previously fixed issue that has resurfaced: python/cpython#61492

---

Update: Relocated code to maintain a continuous cursor busy state during the pymanager-to-interpreter startup sequence.

[zooba]

Good catch. I think we also need to define the variable for the other kind of launcher, too, which lives in _msbuild.py in the top level directory. Want to take a look at that?

[Jesse205]

Sorry, I'm not sure what you're referring to. In pymanager, _WINDOWS only appears here (this is the only location I could find via search).

[zooba]

The _msbuild.py file in the top of the repo is our build script, and it's where Py_WINDOWED is defined. But it's not been defined for the launcherw.exe (template that gets renamed for generating shortcuts), which means these ones won't benefit from this fix.

I can get to it when I get a chance, but it'd be good to fix it all in one PR.

[Jesse205]

launcherw.exe appears to work correctly even without PY_WINDOWED. I just installed Python 3.14 using the released pymanager v26.1 and dragged my .pyzw file onto pythonw3.14.exe. The cursor behaves as expected. Using Process Explorer, I examined the process tree and found that there are no child processes under this pythonw3.14.exe. Additionally, the window it creates is exactly the one generated by my .pyzw file. This seems to differ from how pymanager works.

[zooba]

For runtimes with a compatible python3.dll, yes, we bypass the child process for speed (and can't do it for py.exe because that's already loaded one Python runtime). But if python3.dll is missing, or too old, then it'll launch the executable and we're in the same state.

[zooba]

I'll just take this as it is and file a new issue to fix the other launcher. Thanks for the contribution!

[Jesse205]

This PR fixes a regression where the cursor remains stuck in a busy state.

The _WINDOWS macro is not defined in the pymanager project; it should be PY_WINDOWED. Therefore, I replaced defined(_WINDOWS) with PY_WINDOWED.

This resolves a previously fixed issue that has resurfaced: python/cpython#61492

---

Update: Relocated code to maintain a continuous cursor busy state during the pymanager-to-interpreter startup sequence.

[zooba]

Good catch. I think we also need to define the variable for the other kind of launcher, too, which lives in _msbuild.py in the top level directory. Want to take a look at that?

[Jesse205]

Sorry, I'm not sure what you're referring to. In pymanager, _WINDOWS only appears here (this is the only location I could find via search).

[zooba]

The _msbuild.py file in the top of the repo is our build script, and it's where Py_WINDOWED is defined. But it's not been defined for the launcherw.exe (template that gets renamed for generating shortcuts), which means these ones won't benefit from this fix.

I can get to it when I get a chance, but it'd be good to fix it all in one PR.

[Jesse205]

launcherw.exe appears to work correctly even without PY_WINDOWED. I just installed Python 3.14 using the released pymanager v26.1 and dragged my .pyzw file onto pythonw3.14.exe. The cursor behaves as expected. Using Process Explorer, I examined the process tree and found that there are no child processes under this pythonw3.14.exe. Additionally, the window it creates is exactly the one generated by my .pyzw file. This seems to differ from how pymanager works.

[zooba]

For runtimes with a compatible python3.dll, yes, we bypass the child process for speed (and can't do it for py.exe because that's already loaded one Python runtime). But if python3.dll is missing, or too old, then it'll launch the executable and we're in the same state.

[zooba]

I'll just take this as it is and file a new issue to fix the other launcher. Thanks for the contribution!

[Jesse205]

This PR fixes a regression where the cursor remains stuck in a busy state.

The _WINDOWS macro is not defined in the pymanager project; it should be PY_WINDOWED. Therefore, I replaced defined(_WINDOWS) with PY_WINDOWED.

This resolves a previously fixed issue that has resurfaced: python/cpython#61492

---

Update: Relocated code to maintain a continuous cursor busy state during the pymanager-to-interpreter startup sequence.

[zooba]

Good catch. I think we also need to define the variable for the other kind of launcher, too, which lives in _msbuild.py in the top level directory. Want to take a look at that?

[Jesse205]

Sorry, I'm not sure what you're referring to. In pymanager, _WINDOWS only appears here (this is the only location I could find via search).

[zooba]

The _msbuild.py file in the top of the repo is our build script, and it's where Py_WINDOWED is defined. But it's not been defined for the launcherw.exe (template that gets renamed for generating shortcuts), which means these ones won't benefit from this fix.

I can get to it when I get a chance, but it'd be good to fix it all in one PR.

[Jesse205]

launcherw.exe appears to work correctly even without PY_WINDOWED. I just installed Python 3.14 using the released pymanager v26.1 and dragged my .pyzw file onto pythonw3.14.exe. The cursor behaves as expected. Using Process Explorer, I examined the process tree and found that there are no child processes under this pythonw3.14.exe. Additionally, the window it creates is exactly the one generated by my .pyzw file. This seems to differ from how pymanager works.

[zooba]

For runtimes with a compatible python3.dll, yes, we bypass the child process for speed (and can't do it for py.exe because that's already loaded one Python runtime). But if python3.dll is missing, or too old, then it'll launch the executable and we're in the same state.

[zooba]

I'll just take this as it is and file a new issue to fix the other launcher. Thanks for the contribution!

[Jesse205]

This PR fixes a regression where the cursor remains stuck in a busy state.

The _WINDOWS macro is not defined in the pymanager project; it should be PY_WINDOWED. Therefore, I replaced defined(_WINDOWS) with PY_WINDOWED.

This resolves a previously fixed issue that has resurfaced: python/cpython#61492

---

Update: Relocated code to maintain a continuous cursor busy state during the pymanager-to-interpreter startup sequence.

[zooba]

Good catch. I think we also need to define the variable for the other kind of launcher, too, which lives in _msbuild.py in the top level directory. Want to take a look at that?

[Jesse205]

Sorry, I'm not sure what you're referring to. In pymanager, _WINDOWS only appears here (this is the only location I could find via search).

[zooba]

The _msbuild.py file in the top of the repo is our build script, and it's where Py_WINDOWED is defined. But it's not been defined for the launcherw.exe (template that gets renamed for generating shortcuts), which means these ones won't benefit from this fix.

I can get to it when I get a chance, but it'd be good to fix it all in one PR.

[Jesse205]

launcherw.exe appears to work correctly even without PY_WINDOWED. I just installed Python 3.14 using the released pymanager v26.1 and dragged my .pyzw file onto pythonw3.14.exe. The cursor behaves as expected. Using Process Explorer, I examined the process tree and found that there are no child processes under this pythonw3.14.exe. Additionally, the window it creates is exactly the one generated by my .pyzw file. This seems to differ from how pymanager works.

[zooba]

For runtimes with a compatible python3.dll, yes, we bypass the child process for speed (and can't do it for py.exe because that's already loaded one Python runtime). But if python3.dll is missing, or too old, then it'll launch the executable and we're in the same state.

[zooba]

I'll just take this as it is and file a new issue to fix the other launcher. Thanks for the contribution!