Previously, as soon as an interpreter that wasn't the one having been
used for `env` was used by tox, we would get errors because our C
modules wouldn't be built for this interpreter.
The makefile has been changed to make `make modules` interpreter-aware,
thus fixing this problem.
* Update run.py & .gitignore for windows
- Update run.py to execute on windows as SIGQUIT is not available.
- Update .gitignore to ignore the generate .pyd files
Ref #300, #393
* Update package.py for windows
Add package_windows back into package.py
- Using cx_freeze for freezing installation
- Will be using nsis for actual installer
Tested with python 3.5 64bit on windows 10
Ref #393
* Update makefile for windows (+2 misc)
- Update the makefile to support windows
- Use different bin path in virtualenv
- Use pyd instead of so files
- Tested with Msys2
- Add *.exe to .gitignore
- Fix minor format error in package.py
Ref #393
* Add requirements-windows
Add the requirements-windows.txt
- contains cx-Freeze for bundling
Ref #393
* Add initial setup.nsi
Initial Version of a NSIS installer script
- Multi-user install (install for just one or all)
- Registers uninstaller (more values need to finish up)
- Tested both single and all install / uninstall and works
- Still need to add parameters instead of hardcoded values in some spots
- Need to clean up vendor folders / keys if empty on uninstall
- Need to add the other dupeGuru languages to the language list
- Minor cleanup of script needed as well
Ref #393
* Update setup.nsi
Updates to setup.nsi including:
- Defines from CLI
- Version information (MAJOR, MINOR, PATCH)
- Bits (64 / 32)
- SourcePath (if we wanted something other than build)
- Added extra defines to move application specifics to one location
- Added extra defines for uninstall information
- Added calculation of install size
- Added switching between 64 and 32 bit contexts (need to verify
functionality)
- Updated output file naming
- Added NSIS supported languages which are also supported by dupeGuru
- Added rest of registry keys for uninstall information
- Added missing registry key for installType
- Added removeing Vendor folder and registry key if empty on uninstall
Should be very close to having this installer script ready to integrate
into the package.py script if desired.
Ref #393
* Update README & requirements-windows
Minor update to README to indicate windows is supported. Add PyQt5 to
requirements-windows.txt to make installation easier.
* Update packaging for windows
- Update package.py to integrate NSIS for windows
- Update makefile to deal with a few additional windows issues
- Add Windows.md to contain specific windows instructions, if we want
this can be merged with README.md
- Minor formatting update to setup.nsi
Ref #393
* Update README & Windows Instructions
- Update the README to include a reference to the Windows instructions.
- Add some additional notes into Windows Instructions and remove one
incorrect command.
- Update .gitignore to ignore all permutations of env* to allow for
multiple side by side virtual environments (used to build different
versions for windows)
Ref: #393
* Update Window.md
Fix broken python link and move nsis link for consistency.
* More Details in Windows.md
Update Windows.md including:
- Information on compilier requirements for windows
- Notes about the windows 10 sdk
- Some clarification around some of the steps
- Addition of msys2 links
Going to review this a bit more to polish it up.
Ref #393.
This option allows us to avoid venv+pip-install operations. We can use
this in situations where we already know we have all dependencies met
(in a Gentoo ebuild, for example...) and wish to avoid useless work and
potential problems.
It's a deployment headache. Old sparkle versions generate runtime warnings about security and up to date version requires me to compile on 10.10, but after many tries, it seems that I absolutely need to build on my minimum requirements version which is 10.8. So screw Sparkle.
I finally took the time to properly learn how to write makefiles. This
was long overdue, but here we go.
Much of the makefile wraps `build.py`, but gradually, we'll extract
stuff from there until the makefile is the main container for build
logic.
Ticket #379 reports crashes on quit due to `willSavePrefs` being called
when result and details dialogs are already freed. I can't reproduce the
crash, but it's still a bad idea to rely on the timing of
`aboutToQuit()` to launch this process.
This commits uses a more predictable place to emit `willSavePrefs` and
I'm pretty sure it will fix the crash at #379.
