On Thu, Sep 24, 2020 at 08:38:46AM +0200, Remi Locherer wrote: > On Tue, Sep 08, 2020 at 01:58:09PM +0200, Theo Buehler wrote: > > On Tue, Sep 08, 2020 at 10:33:41AM +0200, Remi Locherer wrote: > > > On Tue, Sep 01, 2020 at 05:59:03PM +0200, Remi Locherer wrote: > > > > This updates rsyslog to version 8.2008.0. The extensive changelog can be > > > > found here: https://github.com/rsyslog/rsyslog/blob/v8.2008.0/ChangeLog > > > > > > > > One notable change is improved support for OpenSSL and also LibreSSL. > > > > But > > > > I did not test that yet. Our port is still using GnuTLS. > > > > > > > > I'm not sure if I should switch rsyslogd from GnuTLS to LibreSSL. I see > > > > the main purpose of rsyslogd on OpenBSD in iterop testing of our > > > > syslogd. For this the use of GnuTLS might be beneficial (except the > > > > slow startup on bluhm's test machine ;-) ). Another usage is writing > > > > messages to different databases or mangling logs. > > > > > > > > Are there any users of this port? If so: how are you using it? > > > > > > > > I tested this update on amd64 with the regress test from rsyslogd. > > > > In addition I verified that the following modules work: > > > > imudp, omrelp (with tls=on), imrelp (with tls=on), ommysql > > > > > > > > OK for the port update? > > > > > > I understand that probably nobody is using rsyslog on OpenBSD. ;-) > > > Nevertheless, can I get an OK? > > > > It would be much nicer to inline the diff rather than quoting it :) > > > > It looks ok portswise and builds fine on amd64 and sparc64. I think > > it'll have better chances of seeing tests if you put this in now rather > > than waiting longer. > > > > ok tb. > > > > There is one somewhat scary warning during the build but it's not new: > > > > imuxsock.c:1112:44: warning: cast from 'char *' to 'struct cmsghdr *' > > increases required > > alignment from 1 to 4 [-Wcast-align] > > for(cm = CMSG_FIRSTHDR(&msgh); cm; cm = > > CMSG_NXTHDR(&msgh, cm)) { > > > > ^~~~~~~~~~~~~~~~~~~~~~ > > /usr/include/sys/socket.h:499:6: note: expanded from macro 'CMSG_NXTHDR' > > (struct cmsghdr *)((char *)(cmsg) + _ALIGN((cmsg)->cmsg_len))) > > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > > I now looked into this one. It is actually not a warning for imuxsock.c but > for OpenBSD's socket.h. > > When adding -Wcast-align to CFLAGS for programs in base, I see many warnings > (checked rad and ospfd). I think this is not something we need to worry > about. But I might be wrong since I'm not an expert. ;-) >
This is the compiler which is not smart enough. cmsg is of type struct cmsghdr * but we need to cast it to char * so that the length can be added to calculate the start of the next header. _ALIGN() will make sure that the result is actually aligned on a long word boundary. So the result is actually 4 / 8 byte aligned and because of this there is no increase of alignment requirement. -- :wq Claudio
