[libvirt] [C# bindings][Pathces] Marshaling corrections, coverage and new sample
Hi there, here are some new patches for the C# bindings. These patches correct marshaling in domain statistics (block / interface / info) and error structure. These parches also add a new sample code. I have included the coverage also. regards, Arnaud
2010/11/13 <arnaud.champion@devatom.fr>:
Hi there,
here are some new patches for the C# bindings. These patches correct marshaling in domain statistics (block / interface / info) and error structure. These parches also add a new sample code. I have included the coverage also.
regards,
Arnaud
Okay, I pushed the 0001-Correct-virGetLastError-marshaling.patch. The 0001-Add-sample-virDomainStats.patch depends on the 0001-Correct-virDomainBlockStats-virDomainInterfaceStats-.patch. In 0001-Correct-virDomainBlockStats-virDomainInterfaceStats-.patch you used UnmanagedType.SysUInt to marshal a ulong in NodeInfo. This fixes the problem with long having different sizes on 32bit and 64bit systems. But for some reason you used UnmanagedType.I8 to marshal long in DomainInterfaceStatsStruct and DomainBlockStatsStruct. I think this is wrong on 32bit systems and should be UnmanagedType.SysInt instead, shouldn't it? Also this comment "// TODO virDomainMemoryStats : Currently disabled in libvirt" is wrong. It's not disabled in libvirt, it's disabled in QEMU. 0001-Forgot-comments-on-new-sample.patch is just an incremental fix for 0001-Add-sample-virDomainStats.patch and I'll squash the fix into it when I can push it. Matthias
Hi Matthias, sorry for the response delay I've been very busy. I was also surprised that this marshaling works without IntPtr, anyway I have double check results against virsh unde Windows/.Net, Windows/Mono, Ubuntu 10.10/Mono (64 bits) and it works fine. Anyway, I can try to put IntPtr to be coherent. What do you think ? Regards, Arnaud -------------------------------------------------- From: "Matthias Bolte" <matthias.bolte@googlemail.com> Sent: Monday, November 15, 2010 5:52 PM To: <arnaud.champion@devatom.fr> Cc: <libvir-list@redhat.com> Subject: Re: [libvirt] [C# bindings][Pathces] Marshaling corrections, coverage and new sample
2010/11/13 <arnaud.champion@devatom.fr>:
Hi there,
here are some new patches for the C# bindings. These patches correct marshaling in domain statistics (block / interface / info) and error structure. These parches also add a new sample code. I have included the coverage also.
regards,
Arnaud
Okay, I pushed the 0001-Correct-virGetLastError-marshaling.patch.
The 0001-Add-sample-virDomainStats.patch depends on the 0001-Correct-virDomainBlockStats-virDomainInterfaceStats-.patch.
In 0001-Correct-virDomainBlockStats-virDomainInterfaceStats-.patch you used UnmanagedType.SysUInt to marshal a ulong in NodeInfo. This fixes the problem with long having different sizes on 32bit and 64bit systems. But for some reason you used UnmanagedType.I8 to marshal long in DomainInterfaceStatsStruct and DomainBlockStatsStruct. I think this is wrong on 32bit systems and should be UnmanagedType.SysInt instead, shouldn't it?
Also this comment "// TODO virDomainMemoryStats : Currently disabled in libvirt" is wrong. It's not disabled in libvirt, it's disabled in QEMU.
0001-Forgot-comments-on-new-sample.patch is just an incremental fix for 0001-Add-sample-virDomainStats.patch and I'll squash the fix into it when I can push it.
Matthias
Okay, you're right I8 is just fine there because DomainInterfaceStatsStruct and DomainBlockStatsStruct just contain long long members and long long is always 64bit width. An int is always 32bit width, so the only thing that needs special handling is the long type in C, as it is compiler and platform dependent if it's 32bit or 64bit width. I applied and pushed the 3 remaining patches now. I also tested them on 32bit Windows and 64bit Linux. Matthias 2010/11/25 <arnaud.champion@devatom.fr>:
Hi Matthias,
sorry for the response delay I've been very busy.
I was also surprised that this marshaling works without IntPtr, anyway I have double check results against virsh unde Windows/.Net, Windows/Mono, Ubuntu 10.10/Mono (64 bits) and it works fine. Anyway, I can try to put IntPtr to be coherent. What do you think ?
Regards,
Arnaud
-------------------------------------------------- From: "Matthias Bolte" <matthias.bolte@googlemail.com> Sent: Monday, November 15, 2010 5:52 PM To: <arnaud.champion@devatom.fr> Cc: <libvir-list@redhat.com> Subject: Re: [libvirt] [C# bindings][Pathces] Marshaling corrections, coverage and new sample
2010/11/13 <arnaud.champion@devatom.fr>:
Hi there,
here are some new patches for the C# bindings. These patches correct marshaling in domain statistics (block / interface / info) and error structure. These parches also add a new sample code. I have included the coverage also.
regards,
Arnaud
Okay, I pushed the 0001-Correct-virGetLastError-marshaling.patch.
The 0001-Add-sample-virDomainStats.patch depends on the 0001-Correct-virDomainBlockStats-virDomainInterfaceStats-.patch.
In 0001-Correct-virDomainBlockStats-virDomainInterfaceStats-.patch you used UnmanagedType.SysUInt to marshal a ulong in NodeInfo. This fixes the problem with long having different sizes on 32bit and 64bit systems. But for some reason you used UnmanagedType.I8 to marshal long in DomainInterfaceStatsStruct and DomainBlockStatsStruct. I think this is wrong on 32bit systems and should be UnmanagedType.SysInt instead, shouldn't it?
Also this comment "// TODO virDomainMemoryStats : Currently disabled in libvirt" is wrong. It's not disabled in libvirt, it's disabled in QEMU.
0001-Forgot-comments-on-new-sample.patch is just an incremental fix for 0001-Add-sample-virDomainStats.patch and I'll squash the fix into it when I can push it.
Matthias
participants (2)
-
arnaud.champion＠devatom.fr -
Matthias Bolte