Updated Windows compatibilty patches

[alex85k]

I have updated Windows compatibility code from
#880
to simplify futher integration.
#646

Many problematic places (like asm code) were already fixed in develop branch, only "big" fixes that remains are UUID renaming, path.c_str() -> path.string().c_str() replacement and xsi shared memory alternative. Tests pass on Windows correctly (when cuke_datastore is not included).

[DennisOSRM]

@alex85k thanks for the updated patch set. Much appreciated! Do you think we should rename the UUID(C) class to give it a better and more telling name?

[alex85k]

I am not very good at naming classes, just tried to make the smallest possible change. :)
Something like OsrmBuildID could work but it is rather long.

[DennisOSRM]

Right, let me meditate a bit on class names. Now, I got to setup the Windows build slave for our Jenkins CI instance.

[alex85k]

Oh, I forgot the updated instructions and batch files: https://gist.github.com/alex85k/8637217#file-a_readme-md
Just tested them on VS2012(need to patch stxxl) and VS2013 (need to patch Boost 1.55).

[DennisOSRM]

@alex85k quick question before I dive into it: Did all the tests pass on Windows?

[alex85k]

Yes, the tests did pass correctly (checked before adding 0624f33 on VS2013 build). Will run recheck to be sure (will post results tomorrow).

[DennisOSRM]

@alex85k awesome. thanks

[alex85k]

I forgot to add - 127.0.0.1 on newer Windows could be much faster than localhost (IPV6 setup problems).

The tests finished, 8 are failing :(

cucumber features\options\files.feature:19 # Scenario: Passing base file
cucumber features\options\help.feature:7 # Scenario: Help should be shown when no options are passed
cucumber features\options\help.feature:30 # Scenario: Help, short
cucumber features\options\help.feature:53 # Scenario: Help, long
cucumber features\options\invalid.feature:7 # Scenario: Non-existing option
cucumber features\options\invalid.feature:14 # Scenario: Missing file
cucumber features\options\version.feature:10 # Scenario: Version, short
cucumber features\options\version.feature:17 # Scenario: Version, long

I guess it is because special color-realted symbols are printed on console.

[alex85k]

Another sample:

  Scenario: Version, long                                                # features\options\version.feature:17
    When I run "osrm-routed --version"                                   # features/step_definitions/options.rb:1
    Then stderr should be empty                                          # features/step_definitions/options.rb:49
      U+043D to US-ASCII in conversion from IBM866 to UTF-8 to US-ASCII (Encoding::UndefinedConversionError)
      ./features/step_definitions/options.rb:50:in `/^stderr should be empty$/'
      features\options\version.feature:19:in `Then stderr should be empty'
    And stdout should contain 1 line                                     # features/step_definitions/options.rb:53
    And stdout should contain /(v\d{1,2}\.\d{1,2}\.\d{1,2}|\w*-\d+-\w+)/ # features/step_definitions/options.rb:35
    And it should exit with code 0                                       # features/step_definitions/options.rb:23

[alex85k]

When running manually, I get

R:\build\11\Project-OSRM\build>osrm-routed.exe  --version
[info] v0.3.7-290-g6f5deb0←[0m

Maybe ←[0m is the problem?

[DennisOSRM]

hm, ok. that's coming from SimpleLogger. I'll look into that. it shouldn't be too hard to fix.

[DennisOSRM]

The commit above should help with the failing tests. @alex85k Can you run them again and report on the status?

[alex85k]

Thank you, the tests are passing after using fixed develop as base brach and pathcing options.rb (adding .exe and quote marks).

273 scenarios (273 passed)
1076 steps (1076 passed)
10m22.433s

(I rebased to develop branch and did git push --force. Hope it will work on Github)

[alex85k]

Gem install bundle failed on Travis. It is not me :)

[emiltin]

looks like it's just a network problem on travis:
Unable to download data from https://rubygems.org/ - Errno::ETIMEDOUT: Connection timed out - connect(2)

[DennisOSRM]

Let me just restart the thing.

[DennisOSRM]

@alex85k passing now.

[alex85k]

Good! If any further testing is needed, let me know.
(updated build instruction: https://gist.github.com/alex85k/8637217#file-a_readme-md )

[DennisOSRM]

Thanks. Will do

[rfrezino]

@alex85k very good! I'm really in troubles trying to compile OSRM for Windows. Do you have some "place" with the binaries ?

Tks

[alex85k]

@rfrezino :
There are "unsupported" non-updating binaries I made for Russian forum:
https://dl.dropboxusercontent.com/u/63393258/OSRM-Windows-038.zip
Pre-windows-7-build(x64):
https://dl.dropboxusercontent.com/u/63393258/OSRM-WindowsXP-038.zip]OSRM-WindowsXP-038.zip

We are waiting for Windows build server and official builds)

[rfrezino]

@alex85k Ok, tks again. I'll keep following your project, it's going to be very useful.

[DennisOSRM]

So, here is the plan to get this into mainline:

• merge other feature branch first
• rebase (then) onto develop branch
• set up Windows VM for CI Server
• configure export of build artifacts
• actually hit merge button

[clns]

After spending a day trying to follow the Windows Compilation guide - with no success in the end - I stumbled upon this. So am I understanding correctly that the guide is outdated, and official builds are going to pop up very soon? I'd like to be able to debug the code.

[alex85k]

@calinseciu Strange, I had updated the guide recently (see also discssion in
http://forum.openstreetmap.org/viewtopic.php?pid=411387#p411387 )
(Wiki guide is outdated)

https://gist.github.com/alex85k/8637217#file-a_readme-md
(it is for release version, however)

[clns]

@alex85k oh I was not aware of this guide. Although no debugging, it should be great to be able to finally compile the project. I'll give it one more try, thanks.

[alex85k]

It can be compile it as debug too (just use debug version of Boost and replace CMAKE_BUILD_TYPE in scripts).

[alex85k]

Let me know when the rebased patch is needed :)

[DennisOSRM]

anytime ;)

[alex85k]

Here it is. Pathches are even smaller now.
(by the way, I have no idea why ca11422 , 1b7a73d , ee7e81f are in this branch)

[DennisOSRM]

Great, can you rebase on develop?

[alex85k]

I did rebase them to develop but pushed them to this pull-request for master. Should I create new PR?

One single test is failing on Windows: features/testbot/oneway.feature ( Tables were not identical (Cucumber::Ast::Table::Different) ). I have no idea why.

[DennisOSRM]

You should be able to just change the target branch.

[alex85k]

As stackoverflow says, I can't :)
http://stackoverflow.com/questions/10081053/how-to-change-the-base-branch-of-a-pull-request

[DennisOSRM]

Alright then lets close here.