Hi! Working on finally getting my group cohomology spkg back to work (at #18514), it seems that internet access during doctests became an issue --- again!
Actually that issue has a long history: - More than three years ago, internet access during doctests was not blocked. So, in my old cohomology spkg, all web-accessing functions used to test (by some kind of introspection) whether they were called as part of a doctest, and would never try to access internet during doctests, UNLESS a special parameter was set to explicitly allow internet access in a few well-chosen tests. - About three years ago, Sage changed so that internet access during doctests was now blocked. Hence, the afore-mentioned well-chosen tests failed, because now no internet access was possible whatsoever. I dealt with it by modifying the well-chosen tests so that web access was simulated by local file access. - Now in SOME very specific configuration (namely: my laptop plugged with an ethernet cable in my office) the test suite hangs, i.e., a wall time of more than 4 hours with a CPU time of 20 minutes. Jeroen seems to be convinced that internet access is to blame. It makes me wonder and gives rise to a whole bunch of questions. 1. Is it really true that Sage currently does NOT block internet access during doctests, i.e., has returned to the situation more than three years ago? If the answer to 1. is "yes", then the follow-up questions are these: 2. Why has Sage returned to allowing internet access? 3. Is there some way (say, by some environment variable) to actively block all internet access while `sage -t` runs? 4. If there isn't: Is it at least possible to block internet access by adding some command at the beginning of each test (say, by setting a timeout in urllib2 to zero or so)? 5. How can one test inside of a python function f whether f was called as part of a doctest? I.e., is the doctesting framework detectable by introspection? To be more specific: By "block internet access" I mean that any attempt to access internet via urllib2 will IMMEDIATELY fail, without a long delay. If the anser to 1. is "no", then the follow-up question is this instead: 6. If internet access is not to blame after all, then what alternative explanation do you have for the fact that THE SAME spkg's test suite on THE SAME laptop works fine when the laptop is connected with WLan at home, bug hangs when the laptop is connected with ethernet cable in my office? Jeroen and John Palmieri did test the spkg at #18514, however it didn't become clear to me whether the tests hang for them. If I understand correctly, Jeroen confirms that internet is accessed during the tests and he can imagine that a failure in accessing the internet may involve a long delay. Best regards, Simon -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at https://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
