| Message ID | 20171213090424.11172-2-yegorslists@googlemail.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [1/4] python-more-itertools: new package | expand |
Hello, On Wed, 13 Dec 2017 10:04:22 +0100, yegorslists@googlemail.com wrote: > Some packages like python-crossbar, python-keyring use scripts, > that check dependencies of all involved packages at runtime. And > though cffi needs pycparser only in host variant during the bindings > compilation, it still has pycparser in its setup.py as dependency. It is not entirely clear to me what is happening here. Could you expand a bit on this? > So always installing pycparser along with cffi would create rather > little overhead, but we won't have to check whether particular > package has such a runtime script, that checks for dependencies. I understand what you mean, but I find the wording confusing. It feels as if you're saying that you're not going to do the change... that you're doing in this commit. What about: "Since adding pycparser as a dependency to python-cffi adds only very little overhead, we chose this solution as it avoids having to ..." Thanks, Thomas
On Sat, Dec 16, 2017 at 4:37 PM, Thomas Petazzoni <thomas.petazzoni@free-electrons.com> wrote: > Hello, > > On Wed, 13 Dec 2017 10:04:22 +0100, yegorslists@googlemail.com wrote: > >> Some packages like python-crossbar, python-keyring use scripts, >> that check dependencies of all involved packages at runtime. And >> though cffi needs pycparser only in host variant during the bindings >> compilation, it still has pycparser in its setup.py as dependency. > > It is not entirely clear to me what is happening here. Could you expand > a bit on this? Let's take python-cryptography, the package using CFFI. But we don't install target pycparser. We can use python-cryptography. But for example keyring has a script named "keyring". And even if you just invoke it with "--help" parameter, it will complain, that you don't have pycparser, because in setuptools "database" cffi package has pycaprser in its install_requires field in setup.py. Actually pycparser is a runtime dependency of cffi [1] and though it is not really needed on target, we have a full moral right to select this package. So I could just change the patch title to "add missing runtime dependency". [1] https://bitbucket.org/cffi/cffi/src/9b653b9a1c876ac94951b980d8dc0ce63dfd0a72/setup.py?at=default&fileviewer=file-view-default#setup.py-214 Yegor >> So always installing pycparser along with cffi would create rather >> little overhead, but we won't have to check whether particular >> package has such a runtime script, that checks for dependencies. > > I understand what you mean, but I find the wording confusing. It feels > as if you're saying that you're not going to do the change... that > you're doing in this commit. > > What about: > > "Since adding pycparser as a dependency to python-cffi adds only very > little overhead, we chose this solution as it avoids having to ..." > > Thanks, > > Thomas > -- > Thomas Petazzoni, CTO, Free Electrons > Embedded Linux and Kernel engineering > http://free-electrons.com
diff --git a/package/python-cffi/Config.in b/package/python-cffi/Config.in index 3805d3d10a..bc9f6a296d 100644 --- a/package/python-cffi/Config.in +++ b/package/python-cffi/Config.in @@ -1,6 +1,7 @@ config BR2_PACKAGE_PYTHON_CFFI bool "python-cffi" select BR2_PACKAGE_LIBFFI + select BR2_PACKAGE_PYTHON_PYCPARSER # runtime help This is the Foreign Function Interface for Python calling C code. The aim of this project is to provide a convenient diff --git a/package/python-crossbar/Config.in b/package/python-crossbar/Config.in index faec922e50..2a98467bd4 100644 --- a/package/python-crossbar/Config.in +++ b/package/python-crossbar/Config.in @@ -5,7 +5,6 @@ config BR2_PACKAGE_PYTHON_CROSSBAR select BR2_PACKAGE_PYTHON_AUTOBAHN select BR2_PACKAGE_PYTHON_CBOR select BR2_PACKAGE_PYTHON_CLICK - select BR2_PACKAGE_PYTHON_PYCPARSER select BR2_PACKAGE_PYTHON_CRYPTOGRAPHY select BR2_PACKAGE_PYTHON_JINJA2 select BR2_PACKAGE_PYTHON_LMDB