[COMMIT osv master] cloud-init: use Linux-compatible syntax for mounts

2017-01-09 Thread Commit Bot
From: Justin Cinkelj Committer: Nadav Har'El Branch: master cloud-init: use Linux-compatible syntax for mounts Previous syntax was for mounts was slightly different as for Linux. In particular, the first entry was tailor made for libnfs

[PATCH] Add feenableexcept() and friends to fenv.h

2017-01-09 Thread Nadav Har'El
Our include/api/fenv.h (copied from Musl) was missing those functions. Signed-off-by: Nadav Har'El --- include/api/fenv.h | 4 1 file changed, 4 insertions(+) diff --git a/include/api/fenv.h b/include/api/fenv.h index 4e31c43..4541513 100644 --- a/include/api/fenv.h +++

[COMMIT osv master] command line: allow --env in runscript

2017-01-09 Thread Commit Bot
From: Justin Cinkelj Committer: Nadav Har'El Branch: master command line: allow --env in runscript If multiple scripts try to set the same environment variable, then the last one wins. Fixes #819 Signed-off-by: Justin Cinkelj

[PATCH 3/4] Implement feenableexcept() and friends

2017-01-09 Thread Nadav Har'El
Implement the functions feenableexcept(), fedisableexcept() and fegetexcept(), non-standard glibc functions which allows controlling which FPU events generate a floating point exception. Signed-off-by: Nadav Har'El --- arch/aarch64/feexcept.cc | 34 ++

[PATCH 1/4] signal: remove "nested signals" protection

2017-01-09 Thread Nadav Har'El
This reverts commit 4af3677117a5bb2f4660d750fa4eddc6ef44e7f2 which aborted on a "nested signal" (handling a signal handler while inside a signal handler). Nested signals were never strictly impossible, but even more importantly, if a signal handler is exited with a longjmp or siglongjmp, our test

[PATCH 2/4] exceptions: support #XM

2017-01-09 Thread Nadav Har'El
The #XM exception ("SIMD exception") may be generated when we request (via feenableexcept()) to trap exceptional FPU cases such as floating point division by zero. We should generate a SIGFPE in this case. This code is currently incomplete in that it doesn't set the "si_code" field for why this

[PATCH 4/4] test for feenableexcept() and friends

2017-01-09 Thread Nadav Har'El
A test for feeenableexcept() and SIGFPE. Works on both Linux and OSv. Signed-off-by: Nadav Har'El --- tests/tst-feexcept.cc | 108 + modules/tests/Makefile | 2 +- 2 files changed, 109 insertions(+), 1 deletion(-) create

Re: Race condition/bug in cloud-init/dhcp?

2017-01-09 Thread Justin Cinkelj
A, and there is at least on case, where wrong transaction ID is normal: [I/42 dhcp]: Waiting for IP... [I/42 dhcp]: Waiting for IP... [W/245 dhcp]: Got packet with wrong transaction ID (1658613712, 83593658) OSv sends first discover. DHCP server takes some time, to check if lease is actually

Re: [PATCH v2] command line: allow --env in runscript

2017-01-09 Thread Nadav Har'El
On Mon, Jan 9, 2017 at 10:06 PM, Justin Cinkelj wrote: > On FC25 test fails. It is compiler and/or std:: lib dependent. > > Diff between v2 nad v3: > -av.reserve(ac+1); > -size_t jj = 0; > -for (auto prm : cmd) { > -av[jj++] =

Re: [PATCH v2] command line: allow --env in runscript

2017-01-09 Thread Justin Cinkelj
On FC25 test fails. It is compiler and/or std:: lib dependent. Diff between v2 nad v3: -av.reserve(ac+1); -size_t jj = 0; -for (auto prm : cmd) { -av[jj++] = prm.c_str(); +av.reserve(ac); +for (auto it=cmd.begin(); it!=cmd.end(); it++) { +

[PATCH v3] command line: allow --env in runscript

2017-01-09 Thread Justin Cinkelj
If multiple scripts try to set the same environment variable, then the last one wins. Fixes #819 Signed-off-by: Justin Cinkelj --- core/commands.cc | 63 +++ tests/tst-commands.cc | 52

Re: [PATCH v2] command line: allow --env in runscript

2017-01-09 Thread Nadav Har'El
Hi Jusin, thanks. I think this version is indeed cleaner. However, I tried: scripts/build image=tests scripts/run.py -e tests/tst-commands.so and I get a failure: ... PASS: runscript multiple lines PASS: runscript multiple lines, multiple commands per line, with args and quotes FAIL: runscript

Re: Race condition/bug in cloud-init/dhcp?

2017-01-09 Thread Justin Cinkelj
The idea of releasing IP during boot was really bad - any code doing TCP/IP during that time could be affected. I saw problem with nfs-mount.so, for example. Sorry for the inconvenience. Justin -- You received this message because you are subscribed to the Google Groups "OSv Development"

[PATCH v2] command line: allow --env in runscript

2017-01-09 Thread Justin Cinkelj
If multiple scripts try to set the same environment variable, then the last one wins. Fixes #819 Signed-off-by: Justin Cinkelj # Conflicts: # loader.cc --- core/commands.cc | 64 +++ tests/tst-commands.cc | 52

Re: [PATCH 2/2] command line: allow --env in runscript

2017-01-09 Thread Justin Cinkelj
Nadav, thank you for looking at it, and for rejecting it. Yes, it simpler to just call boost::program_options directly. Sending v2 patch now. Justin On 01/08/2017 02:00 PM, Nadav Har'El wrote: -- Nadav Har'El n...@scylladb.com On Thu, Dec 29, 2016 at 6:49 PM,