or even use scripting (not fully supported at the moment)I hope you haven't changed any scripting related code.
I hope you haven't changed any scripting related code.
It will make the days of someone pretty funny if you did (the first pushing his stuff wins).
The initial idea was (and still is if i find the time) that you can use generated projects, like from STCube (Microcontroller)This is the job for the generator.
They generate a "generic" makefile you can import in eclipse and i have written a codeblocks import too, but in this file the source files are not listed, only a top level folder is listed. You have to add the whole folder, and every time you regenerate the code you would have to reimport the whole folder and remove the non existent files. Eclipse does this by default.The generators in CMake already do this.
With this function this would happen automatically also in codeblocks.
PS. This was 3 years ago, i do not know if they have updated their project export file to the standard (https://www.keil.com/pack/doc/CMSIS/Build/html/element_cprj.html ) where each file is listed...Even if they have not they would have to do it in order to generate the <glob> stuff.
This is the job for the generator.Well i can not modify the generator....
Does this folder parsing works with multi-folder projects? Would you be able to use these globs with a projects as big as Codeblocks.cbpyes.
If this is the only reason for this feature to exists, I doubt that it is a good idea to add it.This is a example of the usage.... Look at eclipse they use this concept for their projects....
If I were doing it I would monitor the original project definition for changes and just generate plain cbp files. It would be a lot simpler and easier to do. These FS watchers seem scary and inefficient to me and the OS differences are quite massive under the hood...Well i would still need the fs watcher... If you do not use project globs you have absolute no impact (beside the 8 Bytes per project file for the ID)....
I think you have not understood me:I know what it is. I don't know why people want to use it :)
MXCube (a project generator for ST Microcontroller) generates source files and a generic project file (i am not quite sure, but i think it has this format: https://www.keil.com/pack/doc/CMSIS/Build/html/element_cprj.html). This is not a makefile. This is also not a codeblocks project.
If i can not use globs with file system watcher i have to reimport the whole project each time. This is not acceptable. And other IDEs do this automatically. Why should codeblocks not do it?This is what cmake does - it generates a new project file every time.
I understood correctly?No, the UI is in Project->Project Globs
(http://forums.codeblocks.org/index.php?action=dlattach;topic=24297.0;attach=10605)
The initial idea was (and still is if i find the time) that you can use generated projects, like from STCube (Microcontroller)+ :) Now I understand what you offered.
They generate a "generic" makefile you can import in eclipse and i have written a codeblocks import too, but in this file the source files are not listed, only a top level folder is listed. You have to add the whole folder, and every time you regenerate the code you would have to reimport the whole folder and remove the non existent files. Eclipse does this by default.
With this function this would happen automatically also in codeblocks.
PS. This was 3 years ago, i do not know if they have updated their project export file to the standard (https://www.keil.com/pack/doc/CMSIS/Build/html/element_cprj.html (https://www.keil.com/pack/doc/CMSIS/Build/html/element_cprj.html) ) where each file is listed...
BTW:
What is the sales pitch for these globs?
Why would you want to introduce such a complex system?
At work we've used them in cmake and they are a total disaster, so we're slowly replacing them with hard-coded list of files.
I once already experimented with the generated cbp files from cmake, but CB never was able to launch the entire workspace, and after 30 minutes I gave up. Still something I am gonna further investigate.Why are you using the workspace? It should have tons of duplicated stuff. I think cmake generates full projects for every cmake target. Full means all dependent targets are present as C::B targets.
The saving of some settings per "globbed file", might be tricky, and I something I wouldn't do (my feeling : to much added complexity).The behavior has changed from your implementation. In your implementation the glob file was special, you could change the compiler options but they were not saved. This was confusing for me on the first use...