I think that to increase the ability for people to build TBB, we should track the expected hashes in git. I also think that during the ```source-dance```` target, we should ensure that before we unpack anything that we have verified the downloaded files have the expected hashes.
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information
Child items ...
Show closed items
Linked items 0
Link issues together to show that they're related.
Learn more.
So when some component changes, somebody edits the versions.mk file by hand?
That isn't going to work for people who don't like text editors, or when lots of things change. Maybe there should be a script to take a bunch of files and generate that file-with-hashes out of them? Then when there's a new thing, you fetch it, rerun the script, and commit the new file.
And if the versions.mk file does other stuff too, then it may be wisest to keep the auto-generated stuff as a separate file.
I guess it looks like versions.mk had versions in it before, not just hashes, so the problem I describe here might not be a new problem.
So when some component changes, somebody edits the versions.mk file by hand?
That isn't going to work for people who don't like text editors, or when lots of things change. Maybe there should be a script to take a bunch of files and generate that file-with-hashes out of them? Then when there's a new thing, you fetch it, rerun the script, and commit the new file.
The build process is not for people who cannot use a text editor or build software in a terminal.
And if the versions.mk file does other stuff too, then it may be wisest to keep the auto-generated stuff as a separate file.
Currently, it is populated with manual data - eg: version numbers and urls.
I guess it looks like versions.mk had versions in it before, not just hashes, so the problem I describe here might not be a new problem.
No, I don't think so. If we update the version, we should ensure we also either verify the download (and add the hash of the file at the same time) or we should probably not update the version.
(Adding expected hashes when the build scripts fetches a bunch of unauthenticated stuff from the Internet is a great plan.)
OK - glad to hear it. Shall I prepare a patch that checks the hash of the downloaded file against the hash in the Makefile? I'll wait until we hear from Erinn or Mike or both before I spend time implementing such a change.
I've opened #8289 (closed) to address actually using these hashes. This bug should just be about all of the hashes we expect, know and love and not about using the hash for anything.
I think we want to wait to merge #8289 (closed) when we merge this, so we're sure everything is working.
I also think we should only patch the alpha makefiles for now (or at least first?), though. I don't want to risk destabilizing the stable makefiles for this just yet. Of course I also realize that our alpha makefile changes are a pain to backport to the stable makefiles. We need some kind of solution for that (perhaps just copying the alpha makefiles to the stable makefiles at some point?).