Possible Github throttling problem? ("EOF from serverOther error?") #128
Comments
luzpaz
changed the title
|
cc @TheAssassin |
|
I think this error is related to zsync2? in the unofficial one this does not occur(in the lastest master of the git repo), I used In conclusion this is not github's fault @luzpaz . |
|
Weird enough, The CLI tool finishes the update successfully. @luzpaz are you sure this error exists at all? Can you provide the exact AppImage you tried to update? AppImageUpdate seems to work good @probonopd , I think you can close the ticket on zsync2. |
|
@antony-jr I'm sure it exists, I can reproduce it on Manjaro (haven't tried another OS). The AppImage is the one you tested above from: https://github.com/FreeCAD/FreeCAD/releases/tag/0.19_pre |
|
@luzpaz I'm also using Manjaro Linux. I will shortly publish all the test results to confirm this issue. |
|
Okay, Now the AppImage you mentioned is the latest version. Is that the version you've done the update and failed? Because the update will not start if its the latest version. |
|
Ah, It seems you used an old version and I used a latest version. We cannot confirm this unless we have that exact version of AppImage. Is that possible? @luzpaz even if this issue exists I think it will not happen again. It may have occurred because of a network interruption or something else. Have you tried the same thing using the latest version of the AppImage or we could wait for the next release to test this. EDIT: You could use Now update |
|
@antony-jr new appimage available (r17512) from FreeCAD: https://github.com/FreeCAD/FreeCAD/releases/download/0.19_pre/FreeCAD_0.19-17512-Linux-Conda_Py3Qt5_glibc2.12-x86_64.AppImage I'm going to run: on my local (r17499) ResultStalled at Output |
|
The The problem IMO is that the zsync2 code is just not efficient nor fast nor good at all. Therefore there's plans to rewrite the whole project with efficiency and speed as design goals. The whole code just suffers from the old code base we inherited in |
@TheAssassin this type of error for sure doesn't inspire confidence in first-time users, so I'm in agreement (fwiw) |
|
@luzpaz I can confirm this now.
I can help @TheAssassin , We don't really have to write the entire zsync but only the client. You over engineered zsync and implemented everything. We don't want everything all we need is the delta update. (i.e) To do parsing of control file(remotely), calculate and create hash table for seed file, delta writer and block downloader. We don't need control file generation, ZMap or other stuff provided by zsync. We can also use the same for Qt too. |
@TheAssassin the unofficial one seems to EDIT: Sorry wrong conclusion. Needs more testing. |
|
@luzpaz Could you test the same update on https://github.com/antony-jr/AppImageUpdater |
|
|
@antony-jr it looked like it was going to work but then got to 100% without showing it succeeded (offering a call-to-action button or something) it hung there for a long time (longer than I've experienced using the official AppImage-updater (CLI and GUI) then it error'ed out with the following: |
|
Edit: the stdout ends uneventfully: |
@luzpaz Its actually a re-implementation unrelated to AppImageUpdate's code base.
That's on purpose. Anyways, That seems like a network error please retry. Because it used to work on my end. |
That's because the new file is larger than the previous version. Lets just concentrate on why these errors occur. |
|
@antony-jr I deleted the incomplete update from previous attempt and queued the older FC AppImage (r17499). The result is that it just hangs right even before it starts to download anything. The GUI is showing 100% but no data has been downloaded in actuality: |
|
Okay lets stick to the issue, I don't want to debug unrelated application issue here. EDIT : I just updated the program so might be some new bug. |
|
This issue is sooo weird, see antony-jr/AppImageUpdater#16 (comment) , Any ideas @TheAssassin |
|
This might actually be github doing something, like rate limiting high amounts of range requests???? Because this seems to happen when there is large amount of blocks changed?? |
|
Only one way to find out: Put the same files on a different server :-) |
Thanks @probonopd, I tried it in my local server and the error still exists, So this is not github's fault. I think the library needs a big rewrite again.(Both the official and the Qt version) |
|
The internet connection where I usually download the appimage from is pretty slow. I haven't tested what the speed is but I think the error is most likely associated with that |
|
IMO the update should be stable across any internet connection regardless of the speed. |
GitHub is surely rate limiting requests, but as far as I know only for the API, not for the download endpoints. |
|
@TheAssassin I think you can close this. This was fixed long ago. |
DISTRIB_ID=ManjaroLinux
DISTRIB_RELEASE=18.0.4
DISTRIB_CODENAME=Illyria
appimageupdatetool version 1-alpha (commit a62e9b9), build 409 built on 2019-06-15 20:35:29 UTC
Problem:
When updating a FreeCAD appimage using appimagetool, connection will die right before appimage is done updating.
Expecting
That running the appimagetool once will complete the update as expected.
Workaround
Run the same command (below) over again and the AppImage download successfully completes.
Specific Error
Full Output
I can consistently reproduce this. Here is the terminal output:
The text was updated successfully, but these errors were encountered: