> On Aug 28, 2018, at 8:51 AM, Radosław Korzeniewski > <rados...@korzeniewski.net> wrote: > > Hello, > > wt., 28 sie 2018 o 14:42 Dan Langille <d...@langille.org> napisał(a): > The comment in the patch is not accurate: "not defined on BSD systems" > because this is not always true. > > Sorry about that. I'm not a *BSD expert, I wrote my understanding of the > problem. > > > It appears to be a special circumstance as indicated by my initial queries > here: > https://lists.freebsd.org/pipermail/freebsd-ports/2018-August/114171.html > > From the above URL: > > "I'd say it's libc++ missing its errno.h having ENODATA defined if the > following is true: > > - both builds are using clang++ > - both builds are using libc++ > > That header defining ENODATA exists in 11.2 and doesn't exist in 10.4 > (contrib/libc++/include/errno.h)." > > On a FreeBSD 11 box, I get this value: > > $ grep ENODATA /usr/src/contrib/libc++/include/errno.h > #ifndef ENODATA > #define ENODATA 9919 > > Whereas: > $ grep EPIPE /usr/include/c++/v1/errno.h > #ifndef EPIPE > #define EPIPE 9950\ > > On FreeBSD 10.4, we see: > > $ grep EPIPE * > errno.h:#define EPIPE 32 /* Broken pipe */ > > I do not know the implications of using a different value here. Can you > advise here please? > > Well, the fix to this problem is the solution existed in Bacula for a very > long time. I just placed it in one file (bsockcore.c) but forget to add it to > other one (bsock.c). I'm not sure who was an original author for this fix, I > have to check a git log. > > I am still working on finding a better solution. > > Great! Thank you.
I got a reply back from the FreeBSD ports mailing list. I do not yet see that in the mailing list archives, so I will copy/paste it here. The patch has been uploaded to: https://gist.github.com/dlangille/2f59782e4b2ed3bb15922932be8aa0c3 Does this suggestion make more sense than: http://www.bacula.org/git/cgit.cgi/bacula/commit/?h=Branch-9.2&id=eea817f035d32221ab46d1a4f95193cd4d6dbda7 Thank you. ### You can patch the port for now and submit it upstream. Looks like they had ENODATA defined as EPIPE and missed it when they refactored: http://www.bacula.org/git/cgit.cgi/bacula/commit/bacula/src/lib/bsock.c?h=Release-9.2.1&id=c3c124adbe784a55bb31c66401630b2d22768baf. Now that code is in: http://www.bacula.org/git/cgit.cgi/bacula/tree/bacula/src/lib/bsockcore.c?h=Release-9.2.1#n44, but it needs to be in both files. I think it would make more sense to factor it out to src/baconfig.h. That way, they could remove the #if block from bsockcore.c. I've attached a patch and put it in plain text below in case it gets scrubbed. Cheers, Jason Index: sysutils/bacula9-server/files/patch-src_baconfig.h =================================================================== --- sysutils/bacula9-server/files/patch-src_baconfig.h (nonexistent) +++ sysutils/bacula9-server/files/patch-src_baconfig.h (working copy) @@ -0,0 +1,13 @@ +--- src/baconfig.h.orig 2018-08-29 14:09:36 UTC ++++ src/baconfig.h +@@ -38,6 +38,10 @@ + #define have_tls 0 + #endif + ++#ifndef ENODATA ++#define ENODATA EPIPE ++#endif ++ + #ifndef ETIME + #define ETIME ETIMEDOUT + #endif Property changes on: sysutils/bacula9-server/files/patch-src_baconfig.h ___________________________________________________________________ Added: fbsd:nokeywords ## -0,0 +1 ## +yes \ No newline at end of property Added: svn:eol-style ## -0,0 +1 ## +native \ No newline at end of property Added: svn:mime-type ## -0,0 +1 ## +text/plain \ No newline at end of property ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Bacula-devel mailing list Bacula-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-devel
