standardise downloaded tarball location

[ChrisJChang]

Challenge:
Many clusters do not allow us to connect to the internet to download packages, forcing us to have to manually copy them to the cluster.

Idea:
Create a standard location for downloaded tarballs to exist (currently some in Backends/downloaded, ScannerBit/downloaded, inside build folder), and whenever possible, look there first before trying to download anything. This way you could build your branch locally, and then copy a single folder to the cluster. This would avoid a lot of manual copying, and careful checking that something is not forgotten.

An advantage of this is that we could have a nuke mode that specifically did not delete these tarballs (effectively like you did a full nuke-all + recopying the tarballs to the folder).

[andrewfowlie]

Good idea. It may be convenient if this can be controlled by a shell variable eg GAMBIT_DOWNLOADS, and that tarballs were both saved/read from were

…

On 2 Oct 2025, at 9:12 pm, ChrisJChang ***@***.***> wrote:  ChrisJChang created an issue (GambitBSM/gambit#555) Challenge: Many clusters do not allow us to connect to the internet to download packages, forcing us to have to manually copy them to the cluster. Idea: Create a standard location for downloaded tarballs to exist (currently some in Backends/downloaded, ScannerBit/downloaded, inside build folder), and whenever possible, look there first before trying to download anything. This way you could build your branch locally, and then copy a single folder to the cluster. This would avoid a lot of manual copying, and careful checking that something is not forgotten. An advantage of this is that we could have a nuke mode that specifically did not delete these tarballs (effectively like you did a full nuke-all + recopying the tarballs to the folder). — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.

[ChrisJChang]

During the GAMBIT 17 Meeting, we figured out that this might get complicated because it will mess with the CMAKE_BINARY_DIR.

A proposed alternative would be to create a file that contains all of the download paths, and create a script that copies them into the correct locations.