> On June 21, 2017, 9:54 p.m., Jie Yu wrote: > > src/slave/containerizer/mesos/launch.cpp > > Lines 801 (patched) > > <https://reviews.apache.org/r/60280/diff/1/?file=1757586#file1757586line804> > > > > Why this is necessary? I think execvp will look into the current > > working directory? > > Aaron Wood wrote: > I found that this was not true. This is the real core of the fix. It > seems that exec will not look in the current directory, only in PATH.
`The file is sought in the colon-separated list of directory pathnames specified in the PATH environment variable. If this variable isn't defined, the path list defaults to the current directory followed by the list of directories returned by confstr(_CS_PATH). (This confstr(3) call typically returns the value "/bin:/usr/bin".)` - Aaron ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/60280/#review178562 ----------------------------------------------------------- On June 21, 2017, 10:26 p.m., Aaron Wood wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/60280/ > ----------------------------------------------------------- > > (Updated June 21, 2017, 10:26 p.m.) > > > Review request for mesos, Jie Yu, James Peach, and Zhitao Li. > > > Bugs: MESOS-7703 > https://issues.apache.org/jira/browse/MESOS-7703 > > > Repository: mesos > > > Description > ------- > > If a framework specifies use of its own executor and sets shell to false the > executor is never found. Additionally, the name of the binary is never passed > as an argument so executors making use of argv[0] will fail. This provides > the full path to the executor so that the `execvp` or `execvpe` is > successful. The name of the binary is also passed as the first argument for > cases where there is no shell used. > > > Diffs > ----- > > src/slave/containerizer/mesos/launch.cpp 162ca1c2e > > > Diff: https://reviews.apache.org/r/60280/diff/2/ > > > Testing > ------- > > `cd build && cmake .. -DCMAKE_BUILD_TYPE=Release && make -j4` > Also spun up a master and agent, connected and sent a task using the UCR > (both with and without the use of an OCI image) via our own framework, and > checked the sandbox to verify that things went accordingly. > > > Thanks, > > Aaron Wood > >
