Downloads as tar.gz truncated

If I download source code from Github, the downloads vary in size and they are incomplete. Although the webbrowser reports “Completed” which meas :succesful". Up to date Fedora 33 Workstation firefox. Downloads from other websites are OK, this occurs only on GitHub, eiter FreeCAD’s latest tag, or Meshlab’s latest tag.

Hey @atererus and welcome! I’m sorry you’re having issues! Just to confirm, you’re using the Download ZIP button under the Code dropdown on the front page of a repository? Or is this about downloading releases or release artifacts specifically? Could you maybe share a direct link to the asset with us so we can try out ourselves? Thanks!

Hi, thank you for looking into this issue.
I tried several methods and downloads but surely this multiple times : clicking on the small icon “.tar.gz” for Meshlab-2020.12

meanwhile I tried the link :
https:// codeload
with command line wget. This resulted in :
2021-03-28 22:52:33 (61.5 KB/s) - Connection closed at byte 94743721. Retrying.
four times messgaes like 2021-03-29 00:11:23 (62.1 KB/s) - Read error at byte 98266856 (The request is invalid.). Retrying.
And the 6th try resulted in a file about 106MB (never got that far), Unfortunally all files start with wrong data, I asume that I do not use wget right :
wget -l1 --no-parent -A.pdf --continue https : / / github .com/cnr-isti-vclab/meshlab/archive/refs/tags/Meshlab-2020.12.tar.gz"
I’ll try afgain with -o "meshlab.tar.gz: and without the --continue

To me it seems Github connection times out after a specific amount of time, which seems to me 30 minutes. Could you check if there are time limits to download a file on the gitlab file server ?

( Space in links are due to new user restrictions )

Update : the last wget method worked, combined with the fact that some times, the provided speed gets 120KB/s instead of the normal 60KB/s. This got me the 106MB, which is OK. By comparing it with previous downloads, the conclusion is that the data is exactly the same, they ar just truncated mid download.

So I tried the FreeCAD 0.19.1 tag source tar.gz by clicking on the icon, which is bigger in size as the meshlab source. AGain I only get a truncated file at 150 MB, and this is very persistant. Even with the maximum achieved 120KB/s it will not finish within 20 minutes. I then really need either an option to resume a truncated download, or more time in case there exists really some kind of time limit on the github server. Downloads from other websites may break sometimes, but I am able to download a single file for hours and hours. This is not the case for files from GitHub.

Now I now also try to use the wget instead of firefox as in the succesfull meshlab case, the exact syntax is then :
wget -l1 --no-parent -A.pdf “” -O FreeCAD-0.19.1_wget.tar.gz
which failed already 1 time (gzip reports: “unexpected end of file”, and tar extract the source code up to the point of “Unexpected EOF in archive”, which tells me just another case of a partial download).

Well with wget it becomes clear, three times in a row : after 23 minutes, but before 24 minutes after the command was given, a read error occurs. The size of the downloaded just depends on how much it managed to download within that time. The downloaded data is identical up to the end of (truncated) file. As this does not occur with other sites, I guess it must be something github releated. Although I can not find any, even a remote like, complaint about my issue on internet, which strucks me as rather very very odd.

Thereafter it failed 7 times, every time between 23 and 24 minutes, then the failures occured after 36, 26, 35, and 27 minutes overnight. I didn’t get the file, and gave up on it.

PING. I still can not download big source code from GitHub.

Hey @atererus ! I’m so sorry you’re still having this issue! I tried your link above and can download 111 MB of data just fine.

As your investigation uncovers something weird may be happening after 30 minutes, but I don’t have a solid way of verifying this. Note that I am not affiliated so GitHub so sadly I can’t check out any configuration GitHub may or may not have on their side.

Does a git clone work for you? And what happens to be your internet speed?

This is a GitHub problem that is easy to reproduce and breaks products. This command:
wget --limit-rate=100k -O labtainer-master.tar
will fail after about 25 minutes and will begin a retry, which will also fail after about 25 minutes.