Bazel: TensorFlow Build issue with Bazel (Error: cycle in dependency graph)

[srsaharoy]

Hi,

I am opening this as a new issue as suggested by damienmg - the same issue has also been reported as a comment in https://github.com/bazelbuild/bazel/issues/587 and also discussed in https://github.com/bazelbuild/bazel/issues/596

I am building TensorFlow from source. I am using commit 916370e (as suggested in https://github.com/bazelbuild/bazel/issues/596) for Bazel build - Bazel build goes fine.

For building TensorFlow, I had to update the WORKSPACE file (attached) to work-around for known proxy issue (as described in https://github.com/bazelbuild/bazel/issues/587 ) to use "local_repository".

I am using Ubuntu 14.04; From my root (~/tensorflow) I use the following commands to build TensorFlow:

bazel clean --expunge

bazel build -c opt //tensorflow/cc:tutorials_example_trainer

The error that I get from "bazel build ..." command is as follows:

$ bazel build -c opt //tensorflow/cc:tutorials_example_trainer
......
ERROR: /home/ssaharo/.cache/bazel/_bazel_ssaharo/64bf5fcc928610d1fc32c86de697b420/external/png_archive/BUILD:23:1: in genrule rule @png_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:framework
    //tensorflow/core:lib
    //tensorflow/core/platform/default/build_config:platformlib
    @png_archive//:png
  * @png_archive//:configure [self-edge].
ERROR: Loading of target '//tensorflow/cc:tutorials_example_trainer' failed; build aborted.
ERROR: Loading failed; build aborted.
INFO: Elapsed time: 5.852s
============================================

Attached are the WORKSPACE file that I am using, and also png.BUILD and jpeg.BUILD files that I think are somehow related to this issue - both png.BUILD & jpeg.BUILD are original files that come with download of TensorFlow source.

Any help/suggestion would be very much appreciated

Thanks
jpeg.BUILD.txt
png.BUILD.txt
WORKSPACE.txt

[bsilver8192]

I filed a bug for the underlying cause of that failure: #626. As a workaround, png builds for me if I change the exclude line in png.BUILD to exclude = [prefix_dir + "/config.h", "configure"], (probably have to do something similar for the other ones too).

After changing the .BUILD files, make sure to do a bazel shutdown because of #566.

Also, I would put the downloaded folders outside of the Bazel workspace (from your paths, it looks like you have them at the top level of the workspace right now). That worked for me in a simple test, but if you try to build //... or something it will likely not, and it might confuse Bazel about when it needs to rebuild things.

[damienmg]

As said by Ulf the underlying behavior is working as intended. Each file has a corresponding label and you should not name the genrule configure or not include configure in any glob. Is it from TensorFlow source code?

[srsaharoy]

Hi,

yes, all .BUILD files came from TensorFlow Source download

I followed suggestions from bsilver8192:

• Modified the 'exclude' in png.BUILD and jpeg.BUILD as suggested
• Moved the libpng-1.2.53 and jpeg-9a to the directory above 'tensorflow' and marked the path in my WORKSPACE as needed
• did 'bazel shutdown' and started the buid again

Now the build fails while compiling jpeg because it can't find a file in jpeg-9a/jfdctfst.c which does exist.
The error message is as follows:
$ bazel build -c opt //tensorflow/cc:tutorials_example_trainer
......
INFO: Found 1 target...
ERROR: missing input file '@jpeg_archive//:jpeg-9a/jfdctfst.c'.
ERROR: /home/ssaharo/tensorflow/third_party/gpus/cuda/BUILD:143:1: Executing genrule //third_party/gpus/cuda:cuda_config_check failed: bash failed: error executing command /bin/bash -c 'source external/bazel_tools/tools/genrule/genrule-setup.sh; touch bazel-out/local_linux-opt/genfiles/third_party/gpus/cuda/cuda.config': com.google.devtools.build.lib.shell.AbnormalTerminationException: java.io.IOException: Stream closed.
ERROR: /home/ssaharo/.cache/bazel/_bazel_ssaharo/64bf5fcc928610d1fc32c86de697b420/external/jpeg_archive/BUILD:77:1: @jpeg_archive//:jpeg: missing input file '@jpeg_archive//:jpeg-9a/jfdctfst.c'.
Target //tensorflow/cc:tutorials_example_trainer failed to build
Use --verbose_failures to see the command lines of failed build steps.
ERROR: /home/ssaharo/.cache/bazel/_bazel_ssaharo/64bf5fcc928610d1fc32c86de697b420/external/jpeg_archive/BUILD:77:1 1 input file(s) do not exist.
INFO: Elapsed time: 8.098s, Critical Path: 0.11s

I am building CPU version (earlier did _not_ specify GPU while running 'configure' for tensorflow), so I am not sure why the build is executing CUDA build, may be common file!

I am re-attaching the modified WORKSPACE and BUILD files
jpeg.BUILD.txt
png.BUILD.txt
WORKSPACE.txt

[srsaharoy]

by the way, the missing file name changes if I run again after 'bazel shutdown'. As shown below, now its a different file jpeg-9a/jutils.c

$ bazel build -c opt //tensorflow/cc:tutorials_example_trainer
......
INFO: Found 1 target...
ERROR: missing input file '@jpeg_archive//:jpeg-9a/jutils.c'.
ERROR: /home/ssaharo/.cache/bazel/_bazel_ssaharo/64bf5fcc928610d1fc32c86de697b420/external/jpeg_archive/BUILD:77:1: @jpeg_archive//:jpeg: missing input file '@jpeg_archive//:jpeg-9a/jutils.c'.
Target //tensorflow/cc:tutorials_example_trainer failed to build
Use --verbose_failures to see the command lines of failed build steps.
ERROR: /home/ssaharo/.cache/bazel/_bazel_ssaharo/64bf5fcc928610d1fc32c86de697b420/external/jpeg_archive/BUILD:77:1 1 input file(s) do not exist.
INFO: Elapsed time: 7.944s, Critical Path: 0.10s

[bsilver8192]

If you have '/home/ssaharo/jpeg-9a' as the path, then jutils.c should be at /home/ssaharo/jpeg-9a/jpeg-9a/jutils.c (the new_http_archive calls don't use strip_prefix, so the .BUILD files expect that second folder).

Bazel loads the external repositories in parallel, so which one it finds first is not deterministic.

[srsaharoy]

yes, that was it! After moving both libpng-1.2.53 and jpeg-9a directories under /home/ssaharo/libpng-1.2.53 and /home/ssaharo/jpeg-9a,, the build went fine!

Thanks a lot, for your help! This is really great to finally overcome the build issue...

[uriv]

Summary:

• download jpeg.BUILD.txt, png.BUILD.txt, and WORKSPACE.txt from @srsaharoy 's message https://github.com/bazelbuild/bazel/issues/623#issuecomment-158151936 . (the 2nd post with attachments) and place these files in the tensorflow lib without the .txt extension instead of the existing files (in my case ~/git/tensorflow/tensorflow).
• create folder with external source files:
  ~/git/tensorflow/fix/files/re2
  ~/git/tensorflow/fix/files/jpeg-9a/jpeg-9a
  ~/git/tensorflow/fix/files/gemmlowp
  ~/git/tensorflow/fix/files/libpng-1.2.53/libpng-1.2.53
  ~/git/tensorflow/fix/files/six-1.10.0
  Note the dir-in-dir for jpeg-9a and libpng-1.2.53. This is necessary.
• change paths in WORKSPACE file to match the location of the aux source files
• bazel shutdown
• bazel build -c opt //tensorflow/cc:tutorials_example_trainer

EDIT: the path for 'six' in WORKSPACE needs to point to the folder that contains six-1.10.0, e.g.:
path = "/home/.../git/tensorflow/fix/files"
This is because six.BUILD searches the sources in "six-1.10.0/six.py".

EDIT: (for commit a108da2 16-DEC-2015) download & extract https://bitbucket.org/eigen/eigen/get/3.3-beta1.tar.gz ; change eigen_archive to new_local_repository with path = "/home/.../git/tensorflow/fix/files"

This may be also true for other modules.

[damienmg]

Thanks :)

[campuslifeceo]

INFO: Found 1 target...
INFO: From Executing genrule @jpeg_archive//:configure [for host]:
unshare failed with EINVAL even after 101 tries, giving up.
ERROR: /root/.cache/bazel/_bazel_root/73f6290c46c3f436a04f635e253824f5/external/jpeg_archive/BUILD:67:1: Executing genrule @jpeg_archive//:configure failed: bash failed: error executing command
(cd /root/.cache/bazel/_bazel_root/73f6290c46c3f436a04f635e253824f5/tensorflow && \
exec env - \
PATH=/sbin:/usr/sbin:/usr/local/sbin:/root/bin:/usr/local/bin:/usr/bin:/bin:/usr/bin/X11:/usr/X11R6/bin:/usr/games:/usr/lib/mit/bin:/usr/lib/mit/sbin:/root/bin \
/bin/bash -c 'source tools/genrule/genrule-setup.sh; pushd external/jpeg_archive/jpeg-9a; workdir=$(mktemp -d -t tmp.XXXXXXXXXX); cp -a * $workdir; pushd $workdir; ./configure; popd; popd; cp $workdir/jconfig.h bazel-out/host/genfiles/external/jpeg_archive/jpeg-9a; rm -rf $workdir;'): bash failed: error executing command
(cd /root/.cache/bazel/_bazel_root/73f6290c46c3f436a04f635e253824f5/tensorflow && \
exec env - \
PATH=/sbin:/usr/sbin:/usr/local/sbin:/root/bin:/usr/local/bin:/usr/bin:/bin:/usr/bin/X11:/usr/X11R6/bin:/usr/games:/usr/lib/mit/bin:/usr/lib/mit/sbin:/root/bin \
/bin/bash -c 'source tools/genrule/genrule-setup.sh; pushd external/jpeg_archive/jpeg-9a; workdir=$(mktemp -d -t tmp.XXXXXXXXXX); cp -a * $workdir; pushd $workdir; ./configure; popd; popd; cp $workdir/jconfig.h bazel-out/host/genfiles/external/jpeg_archive/jpeg-9a; rm -rf $workdir;').
Target //tensorflow/cc:tutorials_example_trainer failed to build
INFO: Elapsed time: 23.353s, Critical Path: 22.21s

[damienmg]

EINVAL after 101 tries -> set --genrule_strategy=standalone --spawn_strategy=standalone

[uriv]

I've added a small correction for the path to 'six' module in my comment https://github.com/bazelbuild/bazel/issues/623#issuecomment-158984782.

[uriv]

Added another correction for eigen_archive (tensorflow commit a108da2)

[flawless2011]

@uriv Can you elaborate on the eigen changes needed? You mention replacing eigen_archive with a new_local_repository entry. I assume this is in /WORKSPACE? There isn't an eigen_archive entry in that file.

I went ahead and added the new_local_repository entry into WORKSPACE:

...
new_local_repository(
  name = "eigen_archive",
  path = "/Users/me/Dev/tensorflow",
  build_file = "eigen.BUILD"
)
...

With that, it starts building but now get a different error:
ERROR: /Users/me/Dev/tensorflow/tensorflow/tensorflow/core/BUILD:796:1: undeclared inclusion(s) in rule '//tensorflow/core:android_tensorflow_lib': this rule is missing dependency declarations for the following files included by 'tensorflow/core/framework/bfloat16.cc': '/Users/me/Dev/tensorflow/eigen-eigen-ce5a455b34c0/Eigen/src/plugins/BlockMethods.h' '/Users/me/Dev/tensorflow/eigen-eigen-ce5a455b34c0/Eigen/src/plugins/CommonCwiseUnaryOps.h' '/Users/me/Dev/tensorflow/eigen-eigen-ce5a455b34c0/Eigen/src/plugins/CommonCwiseBinaryOps.h' '/Users/me/Dev/tensorflow/eigen-eigen-ce5a455b34c0/Eigen/src/plugins/MatrixCwiseUnaryOps.h' '/Users/me/Dev/tensorflow/eigen-eigen-ce5a455b34c0/Eigen/src/plugins/MatrixCwiseBinaryOps.h' '/Users/me/Dev/tensorflow/eigen-eigen-ce5a455b34c0/Eigen/src/plugins/ArrayCwiseUnaryOps.h' '/Users/me/Dev/tensorflow/eigen-eigen-ce5a455b34c0/Eigen/src/plugins/ArrayCwiseBinaryOps.h'. Target //tensorflow/examples/android:tensorflow_demo failed to build Use --verbose_failures to see the command lines of failed build steps.

Maybe this problem is specific to the android example build.

[uriv]

My WORKSPACE had an "eigen_archive". Here's the fix I made:

#new_http_archive(
#  name = "eigen_archive",
#  url = "https://bitbucket.org/eigen/eigen/get/3.3-beta1.tar.gz",
#  sha256 = "2d6533e86ed6b54d30ae1d6c10808533b335d1c570c5e4c58ce2f03da99c134b",
#  build_file = "eigen.BUILD",
#)

new_local_repository(
  name = "eigen_archive",
  path = "/home/.../git/tensorflow/fix/files",
  build_file = "eigen.BUILD",
)

[flawless2011]

My problem is fixed after pulling the latest source for Tensorflow and Bazel. I think they removed some redundant eigen library files in latest Tensorflow source and that seems to have fixed my issue.

[jeandut]

Hi,
I unfortunately do not have access top internet with my linux computer and I got the same error with missing packages libpng, jpegsrc, re2 and gemmlowp as in #587. I followed @uriv advice and copied and pasted his workpsace png and jpeg build files.
However could someone explain to me very precisely how to do the second step of @uriv summary that is to say creating a local repo with the sources without internet on the computer and changing the path in the corresponding files ? I donwloaded the packages but I do not know where to install them and how to make this work and do not even know if I need the exact version of the packages mentionned.
I am a beginner in Linux and have been trying to install this for a few days and I feel like if I do this it should finally work !

[damienmg]

Bonjour jeandut,
You should download the source in a directory then replace your WORKSPACE file as @uriv commented (with new_local_repository), replacing the path string with the actual path to your source.

[jeandut]

Thank you @damienmg for this quick answer this is what I did and I get:
no such package '@re2//': Could not create a symlink to repository mypath/re2
And the same message for the different other packages it may come from my definition of what a source is. For instance for I just downloaded the zip file for re2 from github the master I unzipped it and renamed the file re2 instead of re2-master is it not the file to which the path must lead ?
Do I need to install it first ? (there is no configure file in re2 files that I got so I am at a loss).
Sorry it is probably a trivial question.

[damienmg]

Can you post a hierarchy of what you did?
It should be something like that:

/path/tensorflow
/path/re2
/path/eigen

then tensorflow/WORKSPACE should look like:

new_local_repository(
  name = "eigen_archive",
  path = "/path/eigen",
  build_file = "eigen.BUILD",
)

local_repository(
  name = "re2",
  path = "/path/re2",
)
...

[jeandut]

I created a folder named fix and inside a folder named files in which I put the unzipped file re2-master. Fix is at the same level as tensorflow and .BUILD files.
So I have
/path/tensorflow
/path/fix/files/re2
etc.
and in the WORKSPACE file I have the same as you do.
Do I need to move them up not to be in subfolders ?
Is it the file re2-master the right one ?

[damienmg]

no the path should not be important as long as they are consistent with the path in the WORKSPACE file.
re2 should be a clone of https://github.com/google/re2

[jeandut]

I cannot even use git in my company so I had to download the zip from the page you are pointing to.
Is that somehow different from cloning from git ?
I am gonna recheck all of the PATH but I am pretty sure they are correct

[damienmg]

no the zip should be good

[jeandut]

It may indeed be a Path mistakeso sorry for that !!!
Could you provide also good link for the other dependencies ?
I only found jpeg-9 instead of 9a and libpng-1.6.20 would that work if I changed the name on WORKSPACE ?
Additionnal question it seems now that I get also an error message for eigen...
I will get back to you as soon as I make progress.

[damienmg]

Don't change the name of the rules in the WORKSPACE file, you can just change the path. It's probably ok with 9a though they haven't tested precisely with those versions.

[jeandut]

I corrected the path for re2 still get the same error message however in the message it displays:
cannot create symlink to repository mypath/re2: /home/jeandut/.cache/bazel/bazel_jeandut/(very long code)/external/re2 (Directory not empty) and referenced by 'tensorflow/core/platform/default/build_config:platformlib
I do not really understand the : and what comes after do you think it is related ?

[damienmg]

try bazel clean --expunge then a build again

On Tue, Jan 12, 2016 at 2:46 PM jeandut [email protected] wrote:

I corrected the path for re2 still get the same error message however in
the message it displays:
cannot create symlink to repository mypath/re2:
/home/jeandut/.cache/bazel/bazel_jeandut/(very long code)/external/re2
(Directory not empty) and referenced by
'tensorflow/core/platform/default/build_config:platformlib

I do not really understand the : and what comes after do you think it is
related ?

—
Reply to this email directly or view it on GitHub
https://github.com/bazelbuild/bazel/issues/623#issuecomment-170915637.

[jeandut]

Yeah !!!!
I got no more mistake for the libraries I installed after entering your command but I still need eigen now I am gonna download it somewhere fairly quickly.
You have been incredibly patient and helpful @damienmg thanks a lot !
I will inform you of my situation as soon as I donwload eigen and copy paste it in the right folder !

[jeandut]

Eigen value works however I encounter another error afterwards coming from the fact that I use libpng-1.2.60. Bad luck ! I will try to find the older version !

[jeandut]

So the step of loading is now fine (thanks again @damienmg). However even after having replaced with the right version libpng and jpeg I got the new message when I build the example:
/home/jeandut/.cache/bazel/_bazel_jeandut/(some long code)/external/jpeg_archive/BUILD:77:1: @jpeg_archive//:jpeg: missing input file @jpeg_archive/jutils.c
And I check inside jepg9-va I got the file ! Any idea ?

[damienmg]

Try bazel clean --expunge

On Tue, Jan 12, 2016, 3:41 PM jeandut [email protected] wrote:

So the step of loading is now fine (thanks again @damienmg
https://github.com/damienmg). However even after having replaced with
the right version libpng and jpeg I got the new message when I build the
example:
/home/jeandut/.cache/bazel/_bazel_jeandut/(some long
code)/external/jpeg_archive/BUILD:77:1: @jpeg_archive//:jpeg: missing input
file @jpeg_archive/jutils.c

And I check inside jepg9-va I got the file ! Any idea ?

—
Reply to this email directly or view it on GitHub
https://github.com/bazelbuild/bazel/issues/623#issuecomment-170932081.

[jeandut]

Unfortunately this magic formula did not work for me a second time !

[damienmg]

What does ls $(bazel info output_base)/external/jpeg_archive/ says?

On Tue, Jan 12, 2016 at 3:48 PM jeandut [email protected] wrote:

Unfortunately this magic formula did not work for me a second time !

—
Reply to this email directly or view it on GitHub
https://github.com/bazelbuild/bazel/issues/623#issuecomment-170936000.

[jeandut]

It is a long list of files including the ones it says are missing.

[jeandut]

Same for png

[damienmg]

Found the problem, there is a prefix_dir there:

https://github.com/tensorflow/tensorflow/blob/master/jpeg.BUILD#L65:
prefix_dir = "jpeg-9a"

You should probably tweak that (that goes for png too).

On Tue, Jan 12, 2016 at 3:53 PM jeandut [email protected] wrote:

It is a long list of files including the ones it says are missing.

—
Reply to this email directly or view it on GitHub
https://github.com/bazelbuild/bazel/issues/623#issuecomment-170937600.

[jeandut]

Thanks d@damienmg but when you say tweak that could you be more precise ?
Shall I instead put the whole path instead ?

[jeandut]

I assume it is related to the dir-in-dir comment but...

[jeandut]

I think I got it I am gonna create a folder containing the folder with the same name is that it ?

[jeandut]

Ok I got it but I get yet another error... I will try to work on it myself not to monopolize the attention !

[jeandut]

I now get the error :
error executing command /usr/bin/gcc -U_FORTIFY_SOURCE -D FORTIFY_SOURCE=1 -fstack- protector -Wall -Wunused-but-set-parameter -Wno-free-nonhea Target //tensorflow/cc:tutorials8example8trainer failed to build
It is frustrating to be so close and yet so far !
I found #591 that says to change something in tools/cpp/CROSSTOOLS however I cannot seem to find it !
Any new idea @damienmg ?

[damienmg]

Can you send the full log? You just show part of the error message.

On Tue, Jan 12, 2016 at 4:56 PM jeandut [email protected] wrote:

I now get the error :
error executing command /usr/bin/gcc -U_FORTIFY_SOURCE -D FORTIFY_SOURCE=1
-fstack- protector -Wall -Wunused-but-set-parameter -Wno-free-nonhea Target
//tensorflow/cc:tutorials8example8trainer failed to build`

It is frustrating to be so close and yet so far !
I found #591 https://github.com/bazelbuild/bazel/issues/591 that says
to change something in CROSSTOOLS however I cannot find it !
Any new idea @damienmg https://github.com/damienmg ?

—
Reply to this email directly or view it on GitHub
https://github.com/bazelbuild/bazel/issues/623#issuecomment-170955826.

[jeandut]

ERROR: /home/jeandut/Bureau/Softwares/TF/tensorflow/tensorflow/core/BUILD:163:1: C++ compilation of rule '//tensorflow/core:framework' failed: gcc failed: error executing command /usr/bin/gcc -U_FORTIFY_SOURCE '-D_FORTIFY_SOURCE=1' -fstack-protector -Wall -Wunused-but-set-parameter -Wno-free-nonheap-object -fno-omit-frame-pointer -g0 -O2 -DNDEBUG -ffunction-sections ... (remaining 69 argument(s) skipped). Target //tensorflow/cc:tutorials_example_trainer failed to build Use --verbose_failures to see the command lines of failed build steps. INFO: Elapsed time: 208.639s, Critical Path: 182.04s. jeandut@jeandut-X7DCA:~/Bureau/Softwares/TF/tensorflow$ bazel clean --expunge INFO: Starting clean (this may take a while). Consider using --expunge_async if the clean takes more than several minutes.

[damienmg]

I meant the lines before...

On Tue, Jan 12, 2016 at 5:05 PM jeandut [email protected] wrote:

ERROR:
/home/jeandut/Bureau/Softwares/TF/tensorflow/tensorflow/core/BUILD:163:1:
C++ compilation of rule '//tensorflow/core:framework' failed: gcc failed:
error executing command /usr/bin/gcc -U_FORTIFY_SOURCE
'-D_FORTIFY_SOURCE=1' -fstack-protector -Wall -Wunused-but-set-parameter
-Wno-free-nonheap-object -fno-omit-frame-pointer -g0 -O2 -DNDEBUG
-ffunction-sections ... (remaining 69 argument(s) skipped).
Target //tensorflow/cc:tutorials_example_trainer failed to build

Use --verbose_failures to see the command lines of failed build steps.

INFO: Elapsed time: 208.639s, Critical Path: 182.04s.
jeandut@jeandut-X7DCA:~/Bureau/Softwares/TF/tensorflow$ bazel clean
--expunge
INFO: Starting clean (this may take a while). Consider using
--expunge_async if the clean takes more than several minutes.

—
Reply to this email directly or view it on GitHub
https://github.com/bazelbuild/bazel/issues/623#issuecomment-170958789.

[jeandut]

Ok sorry I had to rerun it:
log.txt

[damienmg]

the log is still truncated...

On Tue, Jan 12, 2016 at 5:16 PM jeandut [email protected] wrote:

Ok sorry I had to rerun it:
log.txt https://github.com/bazelbuild/bazel/files/87576/log.txt

—
Reply to this email directly or view it on GitHub
https://github.com/bazelbuild/bazel/issues/623#issuecomment-170961600.

[jeandut]

It seems that one of Eigen file is missing unsupported/Eigen/CXX11/Tensor I am gonna reinstall eigen just in case I do not have the right version

[jeandut]

Hopefully I did not leave anything important out this time...
end_of_log.txt

[jeandut]

Could it be because I used eigen 3.2.7 and not the beta version ?

[damienmg]

Yes it is probably. It means that Eigen has not the expected structure.

On Tue, Jan 12, 2016 at 5:28 PM jeandut [email protected] wrote:

Could it be because I used eigen 3.2.7 and not the beta version ?

—
Reply to this email directly or view it on GitHub
https://github.com/bazelbuild/bazel/issues/623#issuecomment-170965189.

[jeandut]

Damn it I cannot test it as the link to the beta is either broken or blocked by my company I will get back home to retry and inform you tomorrow of the result. I really appreciate the effort you put in helping people like me thanks again @damienmg (and Google!).

[jeandut]

Tried with Eigen3 beta version and preexisting version in tensorflow repo get exactly the same error...

[jeandut]

The Tensor file is actually almost empty and only contains the line:

include "external/eigen_archive/eigen-eigen-ce5a455b34c0/unsupported/Eigen/CXX11/Tensor"

furthermore the tensor file used is the one located at third_party which seems weird to me knowing that we also have one in fix/files...
When I go into the directory external/eigen_archive using:
cd $(bazel info output_base)/external/eigen_archive/
the folder eigen-eigen-ce5a455b34c0 does not exist therefore I suppressed it but I still get the same message and it seems to me that it is an infinite loop of including the same file over and over.

[damienmg]

is the path eigen-eigen-ce5a455b34c0 correct? doesn't seems to be

On Tue, Jan 12, 2016 at 6:08 PM jeandut [email protected] wrote:

The Tensor file is actually almost empty and only contains the line:

include

"external/eigen_archive/eigen-eigen-ce5a455b34c0/unsupported/Eigen/CXX11/Tensor"
which might be actually the problem but not sure how to solve it.

—
Reply to this email directly or view it on GitHub
https://github.com/bazelbuild/bazel/issues/623#issuecomment-170977454.

[jeandut]

The cmd bazel build -c opt //tensorflow/cc:tutorials_example_trainer finally ran, thanks to you my versions of eigen were wrong apparently. However during the compilation I got several warnings message about index being out of bond and comparison between signed and unsigned entities.
The example ran but still having problems with gpu support. I opened a new issue (#762 on tensorflow repository) maybe I should move it to bazel ?

[damienmg]

Oh sorry I realized I never give an answer to that last message.

So the GPU thing is probably a TensorFlow side issue. The warnings are, unfortunately, normal.

[jeandut]

Thanks @damienmg ! even late it is still good to know ! Thanks to your advices Tensorflow has been running flawlessly on my computer for a few months !

[damienmg]

Très heureux d'avoir pu aider :)

[shipengfei92]

I met the same issue

[root@node-1 syntaxnet]# bazel test syntaxnet/... util/utf8/...
WARNING: Sandboxed execution is not supported on your system and thus hermeticity of actions cannot be guaranteed. See http://bazel.io/docs/bazel-user-manual.html#sandboxing for more information. You can turn off this warning via --ignore_unsupported_sandboxing.
ERROR: /root/.cache/bazel/_bazel_root/5d85deacc930d23f9397fdea3d7d0f72/external/tf/third_party/eigen3/BUILD:3:1: no such package '@eigen_archive//': Error downloading from https://bitbucket.org/eigen/eigen/get/aaa010b0dd40.tar.gz to /root/.cache/bazel/_bazel_root/5d85deacc930d23f9397fdea3d7d0f72/external/eigen_archive: Error downloading https://bitbucket.org/eigen/eigen/get/aaa010b0dd40.tar.gz to /root/.cache/bazel/_bazel_root/5d85deacc930d23f9397fdea3d7d0f72/external/eigen_archive/aaa010b0dd40.tar.gz: Connection timed out and referenced by '@tf//third_party/eigen3:eigen3'.
ERROR: Loading failed; build aborted.
INFO: Elapsed time: 128.498s
ERROR: Couldn't start the build. Unable to run tests.