Download error for new_http_archive url

[markusdr]

I'm getting the following bazel error with tensorflow:

$ bazel build -c opt --config=cuda //tensorflow/cc:tutorials_example_trainer
ERROR:  tensorflow/tensorflow/core/platform/default/build_config/BUILD:36:1: 
no such package '@jpeg_archive//': 
Error downloading from http://www.ijg.org/files/jpegsrc.v9a.tar.gz

It turns out that bazel cannot download directly from the ijg.org in my network, due to security restrictions. However, I am able to download the file by hand.

I tried just putting it into the required .cache/bazel location by hand and run the bazel command again. This used to work, but it won't work in the latest version, and the same error comes up again.

Would it be possible to enable bazel to pick up a locally downloaded jpegsrc.v9a.tar.gz file? Allowing a local file:// url in new_http_archive should solve the issue.

[ahumesky]

Hi Markus,
You could convert the new_http_archive rule in tensorflow's workspace file
to a new_local_repository rule:
http://bazel.io/docs/be/workspace.html#new_local_repository

Unpack jpegsrc.v9a.tar.gz somewhere and point the new_local_repository to
that path.

On Wed, Jan 27, 2016 at 6:46 PM Markus Dreyer notifications@github.com
wrote:

I'm getting the following bazel error with tensorflow:

$ bazel build -c opt --config=cuda //tensorflow/cc:tutorials_example_trainer
ERROR: tensorflow/tensorflow/core/platform/default/build_config/BUILD:36:1:
no such package '@jpeg_archive//':
Error downloading from http://wwwijgorg/files/jpegsrcv9atargz

It turns out that bazel cannot download directly from the ijgorg in my
network, due to security restrictions However, I am able to download the
file by hand

I tried just putting it into the required cache/bazel location by hand
and run the bazel command again This used to work, but it won't work in the
latest version, and the same error comes up again

Would it be possible to enable bazel to pick up a locally downloaded
jpegsrcv9atargz file? Allowing a local file:// url in new_http_archive
should solve the issue

—
Reply to this email directly or view it on GitHub
#821.

[markusdr]

For some reason, when I convert to new_local_repository it gives another error:

$ bazel build -c opt --config=cuda //tensorflow/cc:tutorials_example_trainer
ERROR: /home/local/ANT/mddreyer/.cache/bazel/_bazel_mddreyer/276918196e903f82b3401cd1a0adce36/external/jpeg_archive/BUILD:67:1: in genrule rule @jpeg_archive//:configure: cycle in dependency graph:
    //tensorflow/cc:tutorials_example_trainer
    //tensorflow/core:tensorflow
    //tensorflow/core:tensorflow_opensource
    //tensorflow/core:core
    //tensorflow/core:core_cpu
    //tensorflow/core:core_cpu_internal
    //tensorflow/core:framework
    //tensorflow/core:framework_internal
    //tensorflow/core:lib_internal
    //tensorflow/core/platform/default/build_config:platformlib
    @jpeg_archive//:jpeg
  * @jpeg_archive//:configure [self-edge].
ERROR: Loading of target '//tensorflow/cc:tutorials_example_trainer' failed; build aborted.

[markusdr]

I figured out a fix for that issue, so new_local_repository is working now. Thanks!

[ahumesky]

Great, glad it's working for you now