Esteban's ChangeLog week of 27 February 2017

Hello!

This is my weekly ChangeLog, from 27 February 2017 to 5 March 2017.
You can see it in a better format by going here: http://log.smallworks.eu/web/search?from=27/2/2017&to=5/3/2017

ChangeLog
=========

3 March 2017:
-------------

*    And now I made +libssh2.so+ compile on linux too :)
   
    I think Iceberg is fixed now for linux, can someone with linux try +vmLatest60+.
   
*    I worked on VM for linux, this time fixing dependencies for libgit2: it was not working because it was trying
    to link libssh2 existing in the system instead the one we provide (a problem with rpath)... I figured out I
    need to add some CMAKE options:
   
    ----
    -DCMAKE_SKIP_BUILD_RPATH=FALSE
    -DCMAKE_INSTALL_RPATH="."
    ----
   
    otherwise +make install+ was ignoring the +-Wl,rpath,.+ configuration. With this it seems to be working,
    but now I have a problem with +libssh2+ compilation.
   
    I'll see this next.
   
*    I took some moments to fix a problem on new UFFI version (0.26): Now, structures use a class variable to keep
    the offset value of a field. So now, instead having an accessor to a field of the form (for example):
   
    ----
    MyStructure>>myField
    ^ handle unsignedLongAt: 42
    ----
   
    ... now you have an accessor of the form:
   
    ----
    MyStructure>>myField
    ^ handle unsignedLongAt: OFFSET_MYFIELD
    ----
   
    this is necesary to allow different platforms to use same structures (because sizes of fields, and in
    consequence of offsets of them inside a stricture are different). This 'offsets' are calculated (as
    structure compiled specs) at first access to +#compiledSpec+ of structure, which usually happens when
    you use it first time... but now always, as [case: 19788](https://pharo.fogbugz.com/f/cases/19788) shows.
   
    Well, now I fixed that :)
   

2 March 2017:
-------------

*    I added a Discord link on [http://pharo.org/community](http://pharo.org/community)... I want to slowly start moving our community there.
    I know Slack is very popular among us this days but Dicords offers some advantages (like no 10k limit on
    messages and a powerful search).
   

1 March 2017:
-------------

*    ... and now I made the [case:19783](https://pharo.fogbugz.com/f/cases/19783), to make SDL2 work on 64bits (it works!) :)
   
*    Yesterday I spent the full day debugging a VM to see why Athens was not working for 64bits and I finally
    figured out the problem was in image and not in VM :P
   
    As a result of the work, I produced this fixes:
   
    * [case:19781](https://pharo.fogbugz.com/f/cases/19781) to fix a problem answering pointers on callbacks.
    * [case:19782](https://pharo.fogbugz.com/f/cases/19782) to actually prepare Athens.
   
    With this two cases, Athens works in 64bits (almost) :)
   
    Now there is a problem in 64bits callbacks that makes the +VGTigerDemo runDemo+ to fail after
    some cycles. Since this problem does not happens on 32bits, I assume there is a problem on how callbacks
    work on 64bits. I'm not working on that (I will work also in migrating SDL2).
   

27 February 2017:
-----------------

*    Well, I spend some time trying to blindly figure out why athens may crash time to time when drawing surfaces,
    basically testing the apprachs [Ronie proposes](http://forum.world.st/Too-frequent-crashes-td4927143i20.html#a4934027).
   
    First approach consists on copying data to not rely on a surface at all. Problem with this is that it involves
    a whole full copy, that causes a lot of GC that pauses a lot the image... so smoothness is gone.
   
    Second approach seems a lot more doable, but I still don't know if it works:
   
    ----
    Form subclass: #AthensCairoSurfaceForm
    instanceVariableNames: 'surface'
    classVariableNames: ''
    package: 'Athens-Cairo'
   
    AthensCairoSurfaceForm>>surface
    ^ surface
   
    AthensCairoSurfaceForm>>surface: anObject
    surface := anObject
   
    AthensCairoSurface>>asForm
    "create a form and copy an image data there"
    self checkSession.
    self flush.
    ^ (AthensCairoSurfaceForm extent: (self width@self height) depth: 32 bits: id)
    surface: self;
    yourself
    ----
   
    this will prevent surfaces to be garbage collected before it's time and seems to be working fine (at least
    in my machine).
   

cheers!
Esteban

btw, this means both Athens and SDL2 should be working on 64bits :)

Hi Esteban,

On Sun, Mar 05, 2017 at 11:00:02PM -0800, [hidden email] wrote:
> *    And now I made +libssh2.so+ compile on linux too :)
>    
>     I think Iceberg is fixed now for linux, can someone with linux try +vmLatest60+.

I've done some quick ad-hoc testing of this and haven't found any
problems (just refreshing, synchronising and pulling incoming commits).


$ ./pharo --version
5.0-201703081637  Wed Mar  8 17:38:04 UTC 2017 gcc 4.6.3 [Production Spur ITHB VM]
CoInterpreter * VMMaker.oscog-eem.2143 uuid: fe064b6b-e530-4766-837d-799ffe1e8dcd Mar  8 2017
StackToRegisterMappingCogit * VMMaker.oscog-eem.2143 uuid: fe064b6b-e530-4766-837d-799ffe1e8dcd Mar  8 2017
VM: 201703081637 https://github.com/pharo-project/pharo-vm.git $ Date: Wed Mar 8 17:37:29 2017 +0100 $
Plugins: 201703081637 https://github.com/pharo-project/pharo-vm.git $
Linux testing-docker-2c74524e-0bce-41a0-905e-309e215345ee 4.8.12-040812-generic #201612020431 SMP Fri Dec 2 09:33:31 UTC 2016 i686 i686 i386 GNU/Linux
plugin path: /home/alistair/pharo6/pharo/pharo-vm/ [default: /home/alistair/pharo6/pharo/pharo-vm/]


HTH,
Alistair

NICE!!