[
https://issues.apache.org/jira/browse/ARROW-12502?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17332514#comment-17332514
]
Roland Weber commented on ARROW-12502:
--------------------------------------
The download logic in the Arrow package on CRAN refers to bintray.
https://github.com/apache/arrow/blob/maint-3.0.x/r/tools/linuxlibs.R#L22
bintray says it's shutting down and will disable all accounts this Saturday,
May 1st.
https://bintray.com/
??Attention: All Bintray services will be deprecated and your accounts will be
disabled on May 1st 2021.??
I guess I'll be looking for an installation path that does not require any
downloads beyond the actual CRAN package. Could it be that the download
problems are because bintray is serving warning pages with the shutdown notice?
Oh... I just noticed that I install first the Arrow package from CRAN, then
pyarrow from PyPI. If I reverse the installation order, maybe the CRAN package
picks up the binary from the pyarrow wheel.
> [R] Download of C++ sources is broken
> -------------------------------------
>
> Key: ARROW-12502
> URL: https://issues.apache.org/jira/browse/ARROW-12502
> Project: Apache Arrow
> Issue Type: Bug
> Components: R
> Affects Versions: 3.0.0
> Reporter: Roland Weber
> Priority: Major
>
> I'm installing Arrow 3.0.0 from CRAN on RedHat UBI. On 2021-04-21, my
> post-installation unit tests for Arrow started to fail. I found this error
> message in the build logs:
> {{*** Successfully retrieved C++ source}}
> {{/bin/gtar: This does not look like a tar archive}}
> {{/bin/gtar: Skipping to next header}}
> {{/bin/gtar: Exiting with failure status due to previous errors}}
> {{*** Proceeding without C++ dependencies}}
> {{Warning message:}}
> {{In untar(tf1, exdir = src_dir) :}}
> {{ ‘/bin/gtar -xf '/tmp/RtmpNhfLVX/file23a66db9da04' -C
> '/tmp/RtmpNhfLVX/file23a640ab53bd'’ returned error code 2}}
> Other installation steps and downloads are working, so I don't think this is
> a network connectivity issue. My guess is that the mirror selection logic
> changed on the server side, so that the source download now saves an HTML
> error page instead of the source archive.
> [https://github.com/apache/arrow/blob/maint-3.0.x/r/tools/linuxlibs.R#L221-L224]
> I'm fixing my build break by switching to binary downloads. But I thought you
> might want to have a look at that source download logic.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
