Has anyone succesfully installed the source gem on 1.9.3dev on any trunk
version?
I just tried installing the gem on trunk@30825 on my Arch Linux system
and got a bunch of errors such as
ragel/redcloth_scan.rl:154:41: error: ISO C90 forbids mixed
declarations and code
I don’t have time to look into it right now or try building from source
on my Win7 box, but it’s likely due to trunk’s use of
-Werror=declaration-after-statement replacing 1.9.2’s
-Wdeclaration-after-statement.
For the record, I think forcing this new restriction downstream is the
wrong thing to do as it breaks extensions that previously worked.
That said, my comments and suggestion when I saw it break the curb gem
didn’t go far:
http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/33974
Can anyone confirm that this is the cause of the build failure I’m
seeing? Assuming it is, how difficult of a task is it to mod all the
code to C90?
Jon
blog: http://jonforums.github.com/
twitter: @jonforums
At 10:02 AM -0500 2/8/11, Jon wrote:
Has anyone succesfully installed the source gem on 1.9.3dev on any trunk version?
I just tried installing the gem on trunk@30825 on my Arch Linux system and got a
bunch of errors such as
ragel/redcloth_scan.rl:154:41: error: ISO C90 forbids mixed declarations and
code
I don’t have time to look into it right now or try building from source on my
Win7 box, but it’s likely due to trunk’s use of
-Werror=declaration-after-statement replacing 1.9.2’s
-Wdeclaration-after-statement.
Jon, did you mean to send this to the ruby-core list?
This is the first I’m aware of it. Thanks for bringing it to the list’s
attention. With any luck, I’ll have re-written RedCloth by the time
this
would be an issue. The way it’s been going, though, maybe not. Please
keep
us posted as this develops in trunk!
Jason
Has anyone succesfully installed the source gem on 1.9.3dev on any trunk
version?
I just tried installing the gem on trunk@30825 on my Arch Linux system and got
a bunch of errors such as
ragel/redcloth_scan.rl:154:41: error: ISO C90 forbids mixed declarations and
code
I don’t have time to look into it right now or try building from source on my
Win7 box, but it’s likely due to trunk’s use of
-Werror=declaration-after-statement replacing 1.9.2’s
-Wdeclaration-after-statement.
Jon, did you mean to send this to the ruby-core list?
No.
I tried that in December 2010 with curb and, while ruby-core was
notified, from that thread it appears there’s no interest on ruby-core’s
part to continue the discussion. If you scan the entire thread, IIRC
Nobu weighed in that he thought the problem was with gem code not
being C90 rather than a problem with trunk build settings.
Fair enough, strongly different PoV’s.
I’m focusing my limited time to provide feedback to the gem authors as
that seems more productive. They can decided what works best for them.
For example, Wayne@FFI decided to tweak his code to be C90.
I still believe it’s a great build setting for enforcing policy for core
Ruby code, but a very bad idea to push it downstream to gem authors.
Gem authors own supporting their code and interaction with their users,
not ruby-core. I think some creative autoconf tweaking could creatively
handle both interests.
Perhaps ruby-core is watching to see if it becomes a real issue by
taking the “don’t ask, just do it and see who squawks…it’s ‘trunk’
after all” strategy. Not a bad idea actually as it tends to sort out
speculation from reality 
Anyways, I had hoped to mock up some autoconf patches to show core, but
I’ve entangled myself into more interesting things.
Jon
blog: http://jonforums.github.com/
twitter: @jonforums