wrong ABI-Version of tensorflow pkg.
Bug #1707002 reported by
Stefan Helmert
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| python-pip (Ubuntu) |
Invalid
|
Undecided
|
Unassigned | ||
Bug Description
On Ubuntu 16.04.2: Installing Tensorflow with pip installs a new tensorflow pkg with old ABI (non C++11).
ProblemType: Bug
DistroRelease: Ubuntu 16.04
Package: python-pip 8.1.1-2ubuntu0.4
ProcVersionSign
Uname: Linux 4.4.0-83-generic x86_64
NonfreeKernelMo
ApportVersion: 2.20.1-0ubuntu2.10
Architecture: amd64
CurrentDesktop: Unity
Date: Thu Jul 27 17:17:53 2017
PackageArchitec
SourcePackage: python-pip
UpgradeStatus: No upgrade log present (probably fresh install)
Revision history for this message
| Stefan Helmert (stefan-helmert) wrote | #1 |
Revision history for this message
| Martin Pecka (peci1) wrote | #2 |
| Changed in python-pip (Ubuntu): | |
| status: | New → Invalid |
To post a comment you must log in.
Why do you think this is a bug? The TF library only links against libpython.so, which is pure C library (so no problem with newer c++11 ABI), a few other pure C system libraries, and against protobuf, which is however built by the same compiler as tensorflow.
Of course, if you want to "misuse" the pip-installed shared objects (I do it, too), it is a problem, but I don't think that's a supported use-case.