[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Tensorflow embedded script usage

I've been looging at tensorflow more seriously.

One bit is confusing me, and I presume you wrote it so I'm wondering
how it is intended to work: there is a debian/embedded.sh script which
unpacks tarballs in debian/embedded dir into external/$foo

This is clearly a mechanism to deal with things that are not yet
packaged in debian. However it's not mentioned in the rules file, so
I'm not quite sure how it is intended to be used.

Is the idea that it is just a manual tool for the maintainer to
temorarily work around missing stuff? Or is it intended to be used by
the rules file for missing/vendored packages/code?

At the moment, if I run it, stuff gets unpacked, but dpkg-source
complains about the large number of new files, some binary, in the
external/ dir (that the embedded script has added). It seems to me
that it would be better if the rules file cleaned this dir and run the
script as a configure stage to set up the source tree cleanly. Did you
just not get round to doing this bit or am I misunderstanding

I've now set it up to work this way and that seems to work neatly.

I'm also a bit confused about the difference between external/ and
third-party/ dirs. I can't see anything that causes the build to look
in the 'external' dir. But my understanding of bazel is very weak so I
could easily be confused.

Cheers for any light you can shed.

Anything else about what you did do, and what you hadn't got round to would also be good.

It seems that currently the tensorflow_framework library (target
//tensorflow:tensorflow_framework ) builds, and so does the c++
library (target //tensorflow:tensorflow_cc), but the c library does
not (complains of missing com_github_grpc_grpc despite libgrpc-dev and
libgrpc++-dev being installed).

Principal hats:  Linaro, Debian, Wookware, ARM

Attachment: signature.asc
Description: PGP signature

Reply to: