[ 
https://issues.apache.org/jira/browse/MESOS-10219?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17364800#comment-17364800
 ] 

acecile5555555 edited comment on MESOS-10219 at 6/17/21, 9:20 AM:
------------------------------------------------------------------

{noformat}
--- a/src/slave/containerizer/mesos/isolators/filesystem/posix.cpp
+++ b/src/slave/containerizer/mesos/isolators/filesystem/posix.cpp
@@ -23,6 +23,7 @@
 #include <stout/fs.hpp>
 #include <stout/os.hpp>
 #include <stout/path.hpp>
+#include <stout/internal/windows/longpath.hpp>

 #include <stout/os/realpath.hpp>

@@ -148,6 +149,7 @@ Future<Nothing> PosixFilesystemIsolatorProcess::update(
   // container paths for volumes. This is enforced by the master. For
   // those volumes, we create symlinks in the executor directory.
   Resources current = info->resources;
+  const std::string directoryPath = info->directory;

   // We first remove unneeded persistent volumes.
   foreach (const Resource& resource, current.persistentVolumes()) {
@@ -168,7 +170,7 @@ Future<Nothing> PosixFilesystemIsolatorProcess::update(
       continue;
     }

-    string link = path::join(info->directory, containerPath);
+    string link = path::join(directoryPath, containerPath);

     LOG(INFO) << "Removing symlink '" << link << "' for persistent volume "
               << resource << " of container " << containerId;
@@ -182,8 +184,9 @@ Future<Nothing> PosixFilesystemIsolatorProcess::update(
   }

   // Get user and group info for this task based on the task's sandbox.
-  struct stat s;
-  if (::stat(info->directory.c_str(), &s) < 0) {
+  struct _stat64i32 s;
+  const std::wstring longDirectoryPath = 
::internal::windows::longpath(directoryPath);
+  if (::_wstat(longDirectoryPath.c_str(), &s) < 0) {
     return Failure("Failed to get ownership for '" + info->directory +
                    "': " + os::strerror(errno));
   }
{noformat}

Came up with this workaround, still have not much idea of what I'm doing but I 
ended up with the following code by matching types according to compiler errors 
:D
Still working fine with short path, does not crash anymore with long path, but 
sadly, the app is still not starting:


{noformat}
Failed to chdir into current working directory 
'C:\mesos\work\slaves\4e393619-8979-47af-8a29-0c6729b5b41a-S1873\frameworks\346d7333-a980-43a8-93ab-343ea12d77d7-0000\executors\tests_windows-azure-vm-with-a-very-very-very-very-very-very-very-very-very-very-very-very-very-long-apth.instance-135dbf9e-c872-11eb-970f-02422c91a06b._app.62\runs\7dfcc8c1-876e-46b5-b79a-9bd24b5c38ce':
 The filename or extension is too long.
{noformat}

I can see the following in app stderr (so I moved one step ahead, mesos is 
actually trying to start the app now).

The code crashing is here:
https://github.com/apache/mesos/blob/master/src/slave/containerizer/mesos/launch.cpp#L1062

So if I understand correctly it should end up here and crashs (if someone can 
confirm I read the code correctly and chdir is actually that call):
https://github.com/apache/mesos/blob/master/3rdparty/stout/include/stout/os/windows/chdir.hpp#L31

But it seems I'm not the only one complaining about SetCurrentDirectoryW 
function:
https://social.msdn.microsoft.com/Forums/vstudio/en-US/fb485167-2093-4465-a916-334882690e19/setcurrentdirectoryw-does-not-work-on-windows-7-for-long-paths?forum=vcgeneral

>From my understanding I should monify a manifest to enable longpathaware:
https://stackoverflow.com/a/44515362
Still trying to figure what this madness is about...


was (Author: acecile5555555):
{noformat}
--- a/src/slave/containerizer/mesos/isolators/filesystem/posix.cpp
+++ b/src/slave/containerizer/mesos/isolators/filesystem/posix.cpp
@@ -23,6 +23,7 @@
 #include <stout/fs.hpp>
 #include <stout/os.hpp>
 #include <stout/path.hpp>
+#include <stout/internal/windows/longpath.hpp>

 #include <stout/os/realpath.hpp>

@@ -148,6 +149,7 @@ Future<Nothing> PosixFilesystemIsolatorProcess::update(
   // container paths for volumes. This is enforced by the master. For
   // those volumes, we create symlinks in the executor directory.
   Resources current = info->resources;
+  const std::string directoryPath = info->directory;

   // We first remove unneeded persistent volumes.
   foreach (const Resource& resource, current.persistentVolumes()) {
@@ -168,7 +170,7 @@ Future<Nothing> PosixFilesystemIsolatorProcess::update(
       continue;
     }

-    string link = path::join(info->directory, containerPath);
+    string link = path::join(directoryPath, containerPath);

     LOG(INFO) << "Removing symlink '" << link << "' for persistent volume "
               << resource << " of container " << containerId;
@@ -182,8 +184,9 @@ Future<Nothing> PosixFilesystemIsolatorProcess::update(
   }

   // Get user and group info for this task based on the task's sandbox.
-  struct stat s;
-  if (::stat(info->directory.c_str(), &s) < 0) {
+  struct _stat64i32 s;
+  const std::wstring longDirectoryPath = 
::internal::windows::longpath(directoryPath);
+  if (::_wstat(longDirectoryPath.c_str(), &s) < 0) {
     return Failure("Failed to get ownership for '" + info->directory +
                    "': " + os::strerror(errno));
   }
{noformat}

Came up with this workaround, still have not much idea of what I'm doing but I 
ended up with the following code by matching types according to compiler errors 
:D
Still working fine with short path, does not crash anymore with long path, but 
sadly, the app is still not starting:


{noformat}
Failed to chdir into current working directory 
'C:\mesos\work\slaves\4e393619-8979-47af-8a29-0c6729b5b41a-S1873\frameworks\346d7333-a980-43a8-93ab-343ea12d77d7-0000\executors\tests_windows-azure-vm-with-a-very-very-very-very-very-very-very-very-very-very-very-very-very-long-apth.instance-135dbf9e-c872-11eb-970f-02422c91a06b._app.62\runs\7dfcc8c1-876e-46b5-b79a-9bd24b5c38ce':
 The filename or extension is too long.
{noformat}

I can see the following in app stderr (so I moved one step ahead, mesos is 
actually trying to start the app now).

The code crashing is here:
https://github.com/apache/mesos/blob/master/src/slave/containerizer/mesos/launch.cpp#L1062

So if I understand correctly it should end up here and crashs (if someone can 
confirm I read the code correctly and chdir is actually that call):
https://github.com/apache/mesos/blob/master/3rdparty/stout/include/stout/os/windows/chdir.hpp#L31

But it seems I'm not the only one complaining about SetCurrentDirectoryW 
function:
https://social.msdn.microsoft.com/Forums/vstudio/en-US/fb485167-2093-4465-a916-334882690e19/setcurrentdirectoryw-does-not-work-on-windows-7-for-long-paths?forum=vcgeneral


> 1.11.0 does not build on Windows
> --------------------------------
>
>                 Key: MESOS-10219
>                 URL: https://issues.apache.org/jira/browse/MESOS-10219
>             Project: Mesos
>          Issue Type: Bug
>          Components: agent, build, cmake
>    Affects Versions: 1.11.0
>            Reporter: acecile5555555
>            Priority: Major
>         Attachments: patch_1.10.0_windows_build.diff
>
>
> Hello,
>  
> I just tried building Mesos 1.11.0 on Windows and this is not working.
>  
> The first issue is libarchive compilation that can be easily workarounded by 
> adding the following hunk to 3rdparty/libarchive-3.3.2.patch:
> {noformat}
> --- a/CMakeLists.txt
> +++ b/CMakeLists.txt
> @@ -137,7 +137,7 @@
>    # This is added into CMAKE_C_FLAGS when CMAKE_BUILD_TYPE is "Debug"
>    # Enable level 4 C4061: The enumerate has no associated handler in a switch
>    #                       statement.
> -  SET(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} /we4061")
> +  #SET(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} /we4061")
>    # Enable level 4 C4254: A larger bit field was assigned to a smaller bit
>    #                       field.
>    SET(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} /we4254")
> {noformat}
> Sadly it is failing later with issue I cannot solve myself:
> {noformat}
> C:\Users\earthlab\mesos\src\csi/state.hpp(22,10): fatal error C1083: Cannot 
> open include file: 'csi/state.pb.h': No such file or directory (compiling 
> source file C:\Users\earthlab\mesos\src\slave\csi_server.cpp) 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>   qos_controller.cpp
>   resource_estimator.cpp
>   slave.cpp
>   state.cpp
>   task_status_update_manager.cpp
>   sandbox.cpp
> C:\Users\earthlab\mesos\src\csi/state.hpp(22,10): fatal error C1083: Cannot 
> open include file: 'csi/state.pb.h': No such file or directory (compiling 
> source file C:\Users\earthlab\mesos\src\slave\slave.cpp) 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>   composing.cpp
>   isolator.cpp
> C:\Users\earthlab\mesos\src\csi/state.hpp(22,10): fatal error C1083: Cannot 
> open include file: 'csi/state.pb.h': No such file or directory (compiling 
> source file C:\Users\earthlab\mesos\src\slave\task_status_update_manager.cpp) 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>   isolator_tracker.cpp
>   launch.cpp
> C:\Users\earthlab\mesos\src\csi/state.hpp(22,10): fatal error C1083: Cannot 
> open include file: 'csi/state.pb.h': No such file or directory (compiling 
> source file C:\Users\earthlab\mesos\src\slave\containerizer\composing.cpp) 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>   launcher.cpp
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launch.cpp(524,34): 
> error C2668: 'os::spawn': ambiguous call to overloaded function 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
> C:\Users\earthlab\mesos\3rdparty\stout\include\stout/os/exec.hpp(52,20): 
> message : could be 'Option<T> os::spawn(const std::string &,const 
> std::vector<std::string,std::allocator<std::string>> &)' 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>           with
>           [
>               T=int
>           ] (compiling source file 
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launch.cpp)
> C:\Users\earthlab\mesos\3rdparty\stout\include\stout/os/windows/exec.hpp(412,20):
>  message : or       'Option<T> os::spawn(const std::string &,const 
> std::vector<std::string,std::allocator<std::string>> &,const 
> Option<std::map<std::string,std::string,std::less<std::string>,std::allocator<std::pair<const
>  std::string,std::string>>>> &)' 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>           with
>           [
>               T=int
>           ] (compiling source file 
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launch.cpp)
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launch.cpp(525,75): 
> message : while trying to match the argument list '(const char [3], 
> initializer list)' [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launch.cpp(893,47): 
> error C2668: 'os::spawn': ambiguous call to overloaded function 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
> C:\Users\earthlab\mesos\3rdparty\stout\include\stout/os/exec.hpp(52,20): 
> message : could be 'Option<T> os::spawn(const std::string &,const 
> std::vector<std::string,std::allocator<std::string>> &)' 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>           with
>           [
>               T=int
>           ] (compiling source file 
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launch.cpp)
> C:\Users\earthlab\mesos\3rdparty\stout\include\stout/os/windows/exec.hpp(412,20):
>  message : or       'Option<T> os::spawn(const std::string &,const 
> std::vector<std::string,std::allocator<std::string>> &,const 
> Option<std::map<std::string,std::string,std::less<std::string>,std::allocator<std::pair<const
>  std::string,std::string>>>> &)' 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>           with
>           [
>               T=int
>           ] (compiling source file 
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launch.cpp)
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launch.cpp(893,47): 
> message : while trying to match the argument list '(const std::string, 
> std::vector<std::string,std::allocator<std::string>>)' 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>   launcher_tracker.cpp
> C:\Users\earthlab\mesos\src\csi/state.hpp(22,10): fatal error C1083: Cannot 
> open include file: 'csi/state.pb.h': No such file or directory (compiling 
> source file 
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launcher.cpp) 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>   mount.cpp
>   switchboard.cpp
> C:\Users\earthlab\mesos\src\csi/state.hpp(22,10): fatal error C1083: Cannot 
> open include file: 'csi/state.pb.h': No such file or directory (compiling 
> source file 
> C:\Users\earthlab\mesos\src\slave\containerizer\mesos\launcher_tracker.cpp) 
> [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>   environment_secret.cpp
>   posix.cpp
>   backend.cpp
>   provisioner.cpp
>   store.cpp
>   authenticatee.cpp
>   authenticator.cpp
>   auxprop.cpp
>   basic_authenticatee.cpp
>   basic_authenticator_factory.cpp
>   combined_authenticator.cpp
>   acls.cpp
>   build.cpp
>   command_utils.cpp
>   protobuf_utils.cpp
>   resource_quantities.cpp
>   resources_utils.cpp
>   roles.cpp
>   type_utils.cpp
>   exec.cpp
>   v0_v1executor.cpp
>   files.cpp
>   hdfs.cpp
>   checker.cpp
>   checker_process.cpp
>   health_checker.cpp
>   devolve.cpp
>   evolve.cpp
>   logging.cpp
>   framework.cpp
>   maintenance.cpp
>   master.cpp
>   quota.cpp
>   quota_handler.cpp
>   readonly_handler.cpp
>   registry_operations.cpp
>   weights.cpp
>   weights_handler.cpp
>   allocator.cpp
>   hierarchical.cpp
>   offer_constraints_filter.cpp
> C:\Users\earthlab\mesos\src\master\allocator\mesos\offer_constraints_filter.cpp(26,10):
>  fatal error C1083: Cannot open include file: 're2/re2.h': No such file or 
> directory [C:\Users\earthlab\mesos\build\src\mesos.vcxproj]
>   messages.cpp
>   daemon.cpp
>   driver.cpp
>   disk_profile_adaptor.cpp
>   sched.cpp
>   scheduler.cpp
>   resolver.cpp
>   in_memory.cpp
>   operation.cpp
>   curl.cpp
>   hadoop.cpp
>   usage.cpp
>   mesos.cpp
>   version.cpp
>   whitelist_watcher.cpp
>   authentication.cpp
>   group.cpp
>   cpu.cpp
>   mem.cpp
>   windows.cpp{noformat}
> If you're interested in getting it built, I'll keep my Windows VM for a 
> moment.
> I'm restarting with older version to see if it goes better.
>  
> Best regards, Adam.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to