This is to audit code and ensure that all error paths are logging errors when failing so that it's easy to see what failed. It looks like at least CFileSystem::Rename doesn't do that, maybe it's more.

in versions.gz, each line contains some info about the version:

byar:git:fff9e7afdb0f45378eadef107c9e5156ae3d32ef,0e0069460c09a7451f162f4927d91340,,Beyond All Reason test-16314-fff9e7a
byar:stable,1f606f2360acdede03e811c5a975b16d,,Beyond All Reason 0.01
byar:test,d78faaf04d16de64ac8c0ba38aa37cde,,Beyond All Reason test-22158-bf74011

Would it be possible to dump this info to stdout? Either before or after a download has finished is fine for my use.

Specifically, the md5 is useful to me in BAR Lobby because the .sdp files in the packages dir use them for filenames, and it helps to be able to know exactly which sdp file is needed ahead of time, instead of iterating through them to search for the version I'm looking for.

i.e. ./pr-downloader.exe --uninstall "Beyond All Reason test-16314-fff9e7a"

I could manually query https://bar-springfiles.p2004a.com/find to get this info, but would prefer to just use prd's --download-map on its own if possible. Just need prd to print the filepath of the downloaded map somewhere so I can cache it

Currently:
./pr-downloader.exe --filesystem-writepath "C:\Users\username\Repos\bar-lobby\dist_electron\content" --validate-sdp "C:\Users\username\Repos\bar-lobby\dist_electron\content\packages\cef6f6184a579cbf25f1dce3c21e305e.sdp"

This has some usage annoyances:

• requires --filesystem-writepath to be specified, otherwise it looks up the related pool files in the wrong dir (defaulting to C:\Users\username\Documents\My Games\Spring\pool\ on windows).
• requires user doing their own version lookup to know which sdp file relates to their content
• is reaaaaally slow for what's essentially meant to be a quick check to see if a package is installed correctly. Would probably require some optimisation/caching to be done on installs to only mark packages as complete after all its files have finished downloading

e.g. instead of ./pr-downloader.exe --download-game byar:git:fff9e7afdb0f45378eadef107c9e5156ae3d32ef, I would like to be able to do ./pr-downloader.exe --download-game "Beyond All Reason test-16314-fff9e7a".

This would save the lobby from having to its own version lookup.

Launched with args:

[
    "C:\\Users\\jaspe\\AppData\\Local\\Beyond All Reason\\Data\\engine\\105.1.1-1158-g27934d7 BAR105\\pr-downloader.exe",
    "--filesystem-writepath",
    "C:\\Users\\jaspe\\AppData\\Local\\Beyond All Reason\\Data",
    "--download-game",
    "Beyond All Reason test-20722-ee934c8"
]

This is a regression from some refactoring a bunch of time ago, and is a safe behavior to just redownload, but it can be optimized to not have to do it.

The implementation would be computing of map file hash on disk, and comparing with the one available in hosting.

Comparing just file presence is not enough, as client could ask to redownload because there is some corruption and existing file on disk is not recognized correctly as valid map.

If-Modified-Since based hashing was disabled in #27 as it was unreliable and racy.

I reported this issue back in April on spring's pr-downloader spring/pr-downloader#145 and it is supposed to have been fixed spring/pr-downloader#145 in there and ported to BAR's pr-downloader as well 34cf14a. However, a user reported the same issue today so it doesn't seem to have been actually fixed.

i.e. ./pr-downloader.exe --list-packages

When the windows installation is fresh (Can be simulated using Windows Sandbox), curl has issue with validating certificate for repos-cdn.beyondallreason.dev domain. Workaround is to load once https://repos-cdn.beyondallreason.dev/ in Edge or Internet Explorer and then certificate validation works just fine.

I believe this is upstream bug in curl: Next steps are to build minimal reproducible example, gather more data, and report upstream.

Task is to extend https://github.com/beyond-all-reason/pr-downloader/blob/master/test/functional_test.py with test for map downloads as currently only game downloads using Rapid are tested.

pr-downloader takes --download-map with the name of map to download (name is SpringName, unique identifier), and calls endpoint defined by PRD_HTTP_SEARCH_URL environment variable to find where it's located and how to download it. E.g.: https://files-cdn.beyondallreason.dev/find?category=map&springname=Angel%20Crossing%201.4 for map with spring name Angel Crossing 1.4.

Parsing in pr-downloader of response is in: https://github.com/beyond-all-reason/pr-downloader/blob/master/src/Downloader/Http/HttpDownloader.cpp#L117

The integration test HTTP server needs to be extended to support such endpoint and serve some small dummy file and test that:

• Download works correctly
• When hash in metadata doesn't match the hash of downloaded file it should error out
• When something fails, it should not create a final .sd7 file but always temp file first

PS: this doesn't require pr-downloader compilation, you can get latest stable binaries from https://github.com/beyond-all-reason/spring-launcher/tree/master/bin and just execute intgration test against them.

This is to help reduce amount of storage/packages that need to be scanned and loaded by spring, by allowing to remove all packages from e.g. byar:* repo, except for some selected tags.

https://discord.com/channels/549281623154229250/773285618826346546/1042241078138507304

This will make it easier for the lobby to parse info from prd. Ideally, each object would follow a standard format, such as:

[Info] /spring/tools/pr-downloader/src/pr-downloader.cpp:178:DownloadSetConfig():Free disk space: 172128 MB

->

{
    "type": "info",
    "message": "Free disk space: 172128 MB"
}

[Progress]   7% [==                            ] 6550225/90399836

->

{
    "type": "download",
    "content": "Beyond All Reason test-16314-fff9e7a",
    "progress": {
        "bytesDownloaded": 6550225,
        "totalBytes": 90399836,
        "percent": 0.07
    }
}

There's also lot of guff in the current stdout that isn't useful for end users, would be preferable to hide that behind a --verbose flag.