> Am 20.06.2017 um 21:56 schrieb Jonathan Corbet <corbet@xxxxxxx>: > > On Tue, 20 Jun 2017 16:51:40 +0200 > Markus Heiser <markus.heiser@xxxxxxxxxxx> wrote: > >> This add a virtualenv [1] with we can control the versions >> of Documentation's (python) requirements. It also changes the >> default behavior, if sphinx is not installed on the OS. >> >> If Sphinx is not available make builds a virtualenv under >> Documentation/local and installs the requirements from >> Documentation/sphinx/requirements.txt with pip [2]. > > I appreciate the work done on this, but I still do not think that going > off and installing external software in response to a kernel make command > is appropriate. :) .. I'm afraid you're right. > Any software installation must require, at a minimum, an > explicit request from the user first. > > It would be nice to have, say, a script that can ensure that a target > system is capable of building the docs. That would preferably include, > as an option, all the awful LaTeX stuff needed for a PDF build. > Preferably it would provide the option of using the system's package > manager or building a venv - but the latter only works for the sphinx > part, of course, and not LaTeX. Hmm, PDF is very hard ... handling package managers from different distribution is also hard ... > I totally agree with taking the pain out of setting up a build > environment, but automatically going off to pull down software off the > net and running it on some developer's build machine is not going to fly. Right, I understand. I did not really think before I send the patch / sorry for wasting your time. First I wan't to focus on python packaging ... Does it make sense if I change the default behavior back to install nothing (bring error messages) and instal only if (e.g.) VENV= has a value? My aim: we all have seen that Sphinx version management is poor. Sphinx only patches on top / not at version level. This is typically for "modern" development where the language brings its own package manager. "Love it or hate it" ... but we should not ignore this fact. To test with different Sphinx and python versions I like to have make options with I can test compilation in different combinations (virtualenv). John, what do you think, does it make sense to continue to work on this patch in that way? -- Markus -- > > Thanks, > > jon -- To unsubscribe from this list: send the line "unsubscribe linux-doc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
