killed tcltest (like for tk) b/c it hangs for hours http11.test make: *** [Makefile:715: test-tcl] Terminated * ERROR: dev-lang/tcl-8.6.6::gentoo failed (test phase): * Make test failed. See above for details. * === This is an issue at stable === ------------------------------------------------------------------- This is an stable amd64 chroot image at a tinderbox (==build bot) name: 13.0-desktop_stable-abi32+64_20170902-123437 ------------------------------------------------------------------- gcc-config -l: [1] x86_64-pc-linux-gnu-5.4.0 * Available Python interpreters, in order of preference: [1] python3.4 [2] python2.7 (fallback) emerge -qpv dev-lang/tcl [ebuild N ] dev-lang/tcl-8.6.6 USE="threads -debug" ABI_X86="32 (64) (-x32)"
Created attachment 491876 [details] emerge-info.txt
Created attachment 491878 [details] dev-lang:tcl-8.6.6:20170902-120858.log
Created attachment 491880 [details] emerge-history.txt
Created attachment 491882 [details] environment
Created attachment 491884 [details] etc.portage.tbz2
Created attachment 491886 [details] logs.tbz2
It looks like the tests are overly restrictive in what they'll accept. They're clearly meant to be ran outside of the package manager. A common problem is "failure in name resolution". Does the tinderbox have a network sandbox, preventing the tests from creating sockets? I'm leaning toward `RESTRICT="test"` since the tests are giving false negatives due to Portage's environment. What are your thoughts, Toralf?
(In reply to Daniel Campbell from comment #7) > It looks like the tests are overly restrictive in what they'll accept. > They're clearly meant to be ran outside of the package manager. A common > problem is "failure in name resolution". Does the tinderbox have a network > sandbox, preventing the tests from creating sockets? > > I'm leaning toward `RESTRICT="test"` since the tests are giving false > negatives due to Portage's environment. > > What are your thoughts, Toralf? I would agree on restricting tests although I am not Toralf. :)
(In reply to Daniel Campbell from comment #7) Well, I run a rather restrictive FEATURES= line: "xattr preserve-libs parallel-fetch ipc-sandbox network-sandbox -news" As far as I understood the devs right, tests cases might be restricted/deactivated b/c the network is restricted by the sandbox.
(In reply to Amy Liffey from comment #8) > I would agree on restricting tests although I am not Toralf. :) Fair point. :) (In reply to Toralf Förster from comment #9) > Well, I run a rather restrictive FEATURES= line: > > "xattr preserve-libs parallel-fetch ipc-sandbox network-sandbox -news" > > As far as I understood the devs right, tests cases might be > restricted/deactivated b/c the network is restricted by the sandbox. Great! I confirmed with a third opinion on IRC, just to be sure. commit 415f646e13d57d37682cdabad76b192af3f4dde0 Author: Daniel Campbell <zlg@gentoo.org> Date: Tue Sep 5 16:19:18 2017 -0700 dev-lang/tcl: restrict faulty testing Testing for this package is designed for use outside of a package manager's environment. FEATURES="network-sandbox" produces hours-long hangs due to incorrect handling of false negatives. Thanks toralf for reporting. Gentoo-Bug: 629680 Gentoo-Bug-URL: https://bugs.gentoo.org/629680 Package-Manager: Portage-2.3.6, Repoman-2.3.3
