Mats Wichmann
Mats Wichmann
The csig is always needed for the sconsign entry so don't see how to avoid it being computed when a target is built. Unfortunately the age of this issue (i.e....
I suppose signature stored as `None`, and if needed the entry would be considered invalid makes at least logical sense, not sure about the code. Added the enhancement tag.
Is there any point in leaving this open? Not much detail, and no action for 20 yes...
This is just weird, why should there be a lexicograhical relationship to the pathname of a tool? Anyway, I try to reproduce it, it doesn't work like this, now if...
Yeah, I was seeing the gcc problem as well, probably for demo purposes should pick a different tool. Also if the action uses `Copy` it should figure out whether mkdir...
You're right, the lex-order-matters thing is still there, I didn't even notice in my test (pasted above) that I dodged it by using `bin/toolname` instead of `tools/toolname`. Just like the...
SCons must be sorting something in order to get reproducibility, and thus somehow introducing this issue?
Attaching a test case made from the original report (suffixed .txt instead of .py due to github limitations) [issue2801.txt](https://github.com/SCons/scons/files/8976804/issue2801.txt)
Isn't there another issue on this topic? The gcc leading-colon syntax is not currently recognized by SCons.
Like make `-o file`, `--old-file=file`