Changes between Version 7 and Version 8 of Building/Preparation/Windows


Ignore:
Timestamp:
Oct 20, 2009 3:00:07 PM (4 years ago)
Author:
JohnD
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • Building/Preparation/Windows

    v7 v8  
    1111I also came across something else that was interesting. Go to [http://www.mingw.org/wiki/LinuxCrossMinGW]. There is a Linux to Windows cross compiler. What does this mean to you and I? In the event that the Windows approach fails you, there has got to be a way. Try OS virtualization, dual-booting your computer, or perhaps better yet pop in a Linux Live CD and attempt to build GHC that way using the MinGW Linux to Windows cross compiler. A USB drive is useful if you pursue the Live CD approach since whatever changes you make to the Linux operating system such as installation of packages will be lost once you restart the computer. If I had known about how GHC could be built by means of the MinGW cross compiler on the front end it would have saved me a few hours. It is time consuming carrying out experimental trials. There is a defect in GHCi that effects me, but I only just been able to build GHC from source. It turns out that in 6.10.4 it was fixed and so in a sense all of this was for nothing. 
    1212 
    13 I encountered some problems attempting to run the test suite. So it may be best to put that off until another day. I have written enough. I don't feel like saying anything more. Suffice it to say I did manage to run both of the test suites, make fast and make stage=2. I couldn't resist now could I? It will be a problem if Windows Firewall is on. Roughly 1% of the tests with "make fast" will fail and took 42 minutes to complete on my computer. "make stage=2" took roughly 6 hours to complete. I thought that before a patch is submitted that it must pass all the make fast tests, at least that was my impression. All the more reason to believe that Windows is not supported in the sense that Linux is. 
     13I encountered some problems attempting to run the test suite. So it may be best to put that off until another day. I have written enough. I don't feel like saying anything more. Suffice it to say I did manage to run both of the test suites, make fast and make stage=2. I couldn't resist now could I? It will be a problem if Windows Firewall is on. Roughly 1% of the tests with "make fast" will fail and took 42 minutes to complete on my computer. "make stage=2" took roughly 6 hours to complete. I thought that before a patch is submitted that it must pass all the make fast tests, at least that was my impression. All the more reason to believe that Windows is not supported in the sense that Linux is. My initial attempts to build the documentation failed, but this doesn't mean that if I were to persist in my effort that I would not find a way. 
    1414 
    1515Now to the symptoms. If you do not apply this fix, all sorts of things will go wrong. I even managed to fix every blessed one of them and was able to successfully build GHC from source, but upon retracing my steps I learned that only one fix was actually needed. Everything else fell into place. The problem is partially described in [http://hackage.haskell.org/trac/GHC/ticket/3201]. It is marked Severity: normal. It is also marked fixed. Fixed? It would warrant, a status of fixed, if someone went through the trouble of writing what I have written out here, but not before. It is like opening a locked door. If you have the key, there is no trouble at all opening the door, but if the key is hidden under a rock a hundred feet away from the house, unless you make it clear to your friend, he or she isn't going to find it. Though I suppose if you brought in a gang of soldiers on their hands and knees like some people suspect occurred in Roswell New Mexico, they might find it. Now try doing this to your girlfriend and call her dumb if he didn't figure it out, "The key was underneath the rock." She might tell you to go crawl under a rock.