On Sat, May 04 2019, Remi Locherer <remi.loche...@relo.ch> wrote: > On Fri, May 03, 2019 at 08:55:40PM +0200, Jeremie Courreges-Anglas wrote: >> On Fri, May 03 2019, Remi Locherer <remi.loche...@relo.ch> wrote: >> > Hi porters >> > >> > This updates librelp to version 1.4.0. It now supports OpenSSL and GNUtls. >> > >> > I verified that rsyslog from ports still works with this (imrelp and >> > omrelp). >> > >> > Please have a look at my patch for tcp.c. Is this a good way to deal with >> > LibreSSL not supporting @SECLEVEL? >> >> I think it's fine. No need to spread the >> "!defined(LIBRESSL_VERSION_NUMBER)" disease. >> >> > Comments/OKs? >> >> As noted by Antoine the python version shouldn't be hardcoded. Here's >> a proposal which: >> - removes gnutls from TEST_DEPENDS (LIB_DEPENDS is enough) >> - adds coreutils to TEST_DEPENDS for gtimeout, so that the *abort* tests >> aren't skipped >> - uses MODPY_RUN_DEPENDS and MODPY_BIN from the python module instead of >> hardcoding the python path and version. We could have used >> MODPY_ADJ_FILES=tests/*.py but tests/test-framework.sh would have >> needed patching anyway. >> - use python2, needed by tests/dummyserver.py (previously SKIPPED >> because of the lack of "timeout" command) >> - moves the /bin/bash handling to post-extract so that update-patches >> doesn't create bogus patches >> >> Note that you could also add a TEST_DEPENDS on sysutils/ggrep instead of >> patching tests/tls-wrong-signedcert.sh. >> >> The major bump is warranted (symbols have been removed) and the update >> looks good ports-wise. > > Thanks a lot for all these inputs! > > I like your suggestion to add ggrep to TEST_DEPENDS. I believe that lowers > future maintenance work.
Yep. Since you have tested consumers, ok jca@ -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE