with 4.9.5 there came a new vfs module glusterfs_fuse, which breaks the build on some platforms, just experienced this on AIX. The gluster module use the NAME_MAX macro, which may NOT nbe expected to be defined by POSIX, see https://pubs.opengroup.org/onlinepubs/7908799/xsh/limits.h.html, you should use pathconf() to get the value of NAME_MAX using that. I'm also wondering if a the glusterfs modules shouldn't be only built on platforms there glusterfs is actually availble on.
ping
Created attachment 15074 [details] patch from master for v4.9 and v4.10
Karolin, please add the patches to the relevant branches. Thanks!
Pushed to autobuild-v4-{10,9}-test.
this patch defies NAME_MAX statucally to 255 in replace.h. This is the wrind approach. My initial comment already mentioned the corrwct approach, please undo the static setting to 255
in addition to that please do not add this module to the default modules to build for all (also glusterfs unsupported) platforms
(In reply to Björn Jacke from comment #5) I agree, replace.h is entirely the wrong way to solve a missing define like this. If this were in something glusterfs only, that might make more sense, but then one would look at the actual code and wonder why NAME_MAX is in use at all. Samba got rid of fstring deliberately, this is just brining it back, badly. Guenther, The quick way out of this is to just move that to the top of the modules, defining GLUSTER_NAME_MAX to NAME_MAX or 255, but for master, please rework the code to use a dynamically allocated buffer, eg talloc_asprintf() on the input and either pathconf() or an ERANGE detecting loop on the output side.
ping This currently blocks 4.9.7...
Created attachment 15099 [details] patch from master for v4.9 and v4.10
(In reply to Guenther Deschner from comment #9) Pushed to autobuild-v4-{10,9}-test. Thanks!
(In reply to Karolin Seeger from comment #10) Pushed to both branches, thanks! Re-assigning to Günther due to Björn's request not to build on systems that do not support gluster.
Assigning to Karolin for the next release.
Created attachment 15250 [details] Followup patch from master for 4.10 and 4.9 Following Volker's advice off-list, follow-up patch (reverting the old one) using a simple re-define has been pushed upstream. The old patch was also not working as pathconf() did not deal well with "." directory causing both vfs modules not to operate correctly at all.
(In reply to Guenther Deschner from comment #13) Pushed to autobuild-v4-{10,9}-test.
(In reply to Karolin Seeger from comment #14) Pushed to both branches. Closing out bug report. Thanks!