I've got filesystems hosted on centos 7. They are mounted (nfs4) on my newly configured fedora 29 desktop.
I'm debugging an executable that lives on the NFS filesystem. I make changes and recompile over on the centos system.
I try testing the newly compiled executable on fedora 29 and it acts totally strange in ways that make no sense.
I reboot the fedora 29 system and try testing again, and everything seems to be as expected.
I can't put my finger on anything specific, but it sure seems like NFS is at fault somehow.
When I had fedora 28 on this same desktop and was doing the same kind of development, the strangeness never manifested.
On 11/6/18 6:57 AM, Tom Horsley wrote:
I've got filesystems hosted on centos 7. They are mounted (nfs4) on my newly configured fedora 29 desktop.
I'm debugging an executable that lives on the NFS filesystem. I make changes and recompile over on the centos system.
I try testing the newly compiled executable on fedora 29 and it acts totally strange in ways that make no sense.
That's singularly useless in letting us help you. Be specific as to what didn't work.
I reboot the fedora 29 system and try testing again, and everything seems to be as expected.
I can't put my finger on anything specific, but it sure seems like NFS is at fault somehow.
Try turning off caching on the NFS mount by specifying "noac" in the options. I don't know if there's a different default for F29 versus F28, but caching can confuse things. I/O will be slower, but things will be coherent. ---------------------------------------------------------------------- - Rick Stevens, Systems Engineer, AllDigital ricks@alldigital.com - - AIM/Skype: therps2 ICQ: 226437340 Yahoo: origrps2 - - - - Admitting you have a problem is the first step toward getting - - medicated for it. -- Jim Evarts (http://www.TopFive.com) - ----------------------------------------------------------------------
On Tue, 6 Nov 2018 17:46:58 +0000 Rick Stevens wrote:
I try testing the newly compiled executable on fedora 29 and it acts totally strange in ways that make no sense.
That's singularly useless in letting us help you. Be specific as to what didn't work.
The behavior is singularly random, so it is hard to describe usefully. It is like the program was designed to compute PI to 10 digits, but instead it draws a picture of a duck :-).
I'm beginning to suspect that my actual problem is that the sigsetops routines have changed in the new glibc and I'm getting random trash in signal sets because I'm getting errors from things like sigaddset I didn't get with f28 libs and I'm going down strange and wondrous code paths.
I still have no idea why the behavior always radically changes when I relink the program on the remote system though.
On 06Nov2018 09:57, Tom Horsley horsley1953@gmail.com wrote:
I've got filesystems hosted on centos 7. They are mounted (nfs4) on my newly configured fedora 29 desktop.
I'm debugging an executable that lives on the NFS filesystem. I make changes and recompile over on the centos system.
I try testing the newly compiled executable on fedora 29 and it acts totally strange in ways that make no sense.
Compiled on the Centos system, but dynamicly linking to library from the Fedora system? That could be a bad combination.
Are the outputs of "ldd your-executable", run on fedora and centos, different?
I reboot the fedora 29 system and try testing again, and everything seems to be as expected.
Weird.
I can't put my finger on anything specific, but it sure seems like NFS is at fault somehow.
So: copy the source to fedora and compile on fedora. Is the behaviour normal or weird?
Guessing in the dark here, but a dynamicly linked executable will be getting its library from the local system. But if it was compiled remotely the headers will describe the remote system libraries. Binary level badness if things mismatch, possibly.
Cheers, Cameron Simpson cs@cskk.id.au