sanitize interrupt logging, bump compiler, merge upstream, build simulator, snapshot fixes#11
Conversation
2a83977 to
14572af
Compare
7e42ef9 to
5a361be
Compare
Nieuwejaar
left a comment
There was a problem hiding this comment.
A couple of minor nits, but this looks great. I'm shocked that it took so little to get the model to build.
What happened with boost? I forget why I did all that in the first place. Was it not in helios at all? Did helios have an old version? Wait...did I add it to helios?
| phv_recs++; | ||
| continue; | ||
| } | ||
| /* Only use the record with field_lsb == 0 (the primary record) */ |
There was a problem hiding this comment.
"Record" or "records"? If there is only one record in the snapshot we care about, shouldn't we be breaking out of the loop after the memcpy below? If there is more than one "primary" record, when (if ever) do the secondary records get consumed?
Oh, I see now. We do break out of the loop, but keyed on field_found. That makes sense, just to keep the same structure as the loop below, but a comment at the top of the loop would help.
|
|
||
| #else | ||
|
|
||
| // TODO: emulate these if building for non-Linux OS |
There was a problem hiding this comment.
Interesting. I wonder what non-Linux OS they were using. FreeBSD?
| @@ -0,0 +1,65 @@ | |||
| /* | |||
| * Stub implementation of portmanager for illumos. | |||
| * Real network interface support would require DLPI implementation. | |||
There was a problem hiding this comment.
What does this mean? If the model was trying to use a nic directly, rather than a TCP socket?
There was a problem hiding this comment.
This is for the simulated port interfaces that plug into the model. The thing that is done with vnic devices on linux.
Pulling in the latest p4c brought in updates to the c++ standard being used as well as the version of the boost the compiler was using. Compiling the SDE with one version of c++/boost and p4c with another was a real mess, so I just elected to use the system's boost on both linux and helios which is much newer than the vendored boost and works fine with c++20. |
…ckage (p4lang#85) Signed-off-by: Andy Fingerhut <andy_fingerhut@alum.wustl.edu>
Signed-off-by: fruffy <fruffy@nyu.edu>
Signed-off-by: fruffy <fruffy@nyu.edu> Signed-off-by: Fabian Ruffy <5960321+fruffy@users.noreply.github.com>
Signed-off-by: Andy Fingerhut <andy_fingerhut@alum.wustl.edu>
…paces with temporary directories. (p4lang#94) Signed-off-by: fruffy <fruffy@nyu.edu>
…4lang#65) Signed-off-by: fruffy <fruffy@nyu.edu> Co-authored-by: Prathima Kotikalapudi <prathima.kotikalapudi@intel.com>
…port (p4lang#84) Signed-off-by: Andy Fingerhut <andy.fingerhut@gmail.com>
Signed-off-by: fruffy <fruffy@nyu.edu>
Signed-off-by: Vladimir Gurevich <vag@p4ica.com> Signed-off-by: fruffy <fruffy@nyu.edu>
Signed-off-by: Andy Fingerhut <andy.fingerhut@gmail.com> Signed-off-by: Fabian Ihle <fabian.ihle@uni-tuebingen.de> Signed-off-by: fruffy <fruffy@nyu.edu> Signed-off-by: Vladimir Gurevich <vag@p4ica.com> Co-authored-by: Andy Fingerhut <andy.fingerhut@gmail.com> Co-authored-by: Fabian Ruffy <5960321+fruffy@users.noreply.github.com> Co-authored-by: Vladimir Gurevich <vag@p4ica.com>
Signed-off-by: sai-pracheetha <sai.beeyam@intel.com>
Signed-off-by: Alexander Gall <alexander.gall@gmail.com>
Signed-off-by: fruffy <fruffy@nyu.edu>
1592b7e to
9e7c601
Compare
Interrupts were being logged to a buffer that only shows up when the owning process exists. Change interrupt logging to use the same macros as everybody else so log entries show up in the overall dpd log in a reasonable way.
Add support for 128 bit snapshot targets and fix upper bound range bug in stage selection for snapshots.
Merge from upstream. Notably, this pulls in the tofino model simulator.
Depends on merge from upstream tofino-p4c#2
The vast majority of this PR is an upstream merge. My actual changes begin at fd6a6fd.