Source Location of DataCon objects in GHC 7.4.1 API

[Jacques Carette]

One bit of experience from my Maple days: we made it a policy to create 
regression tests for every single bug we fixed.  No bug could be marked 
closed without an identifier for the test that makes sure this does not 
revert.

This seemed like too much at first, but over the years, we were always 
glad of this.  We caught all sorts of very strange bugs because of doing 
this.  This is especially valuable for very long-lived software (like 
GHC), where the time-span between the last time the problem showed up 
and when it occurs again can be 15+ years.

In other words, I am saying that even though in the present context this 
feature/bug can seem small, it is extremely hard to predict what context 
will be true in 10+ years, where this regression may become a symptom of 
something bigger.

Jacques

On 05/06/2012 8:36 AM, Simon Peyton-Jones wrote:
> Done.  I don't think it's worth a regression test.
>
> SImon
>
> commit cb705a38d677e2ab4cad37447c8180bd397d5576
> Author: Simon Peyton Jones<simonpj at microsoft.com>
> Date:   Tue Jun 5 13:35:07 2012 +0100
>
>      Add sensible locations to record-selector bindings
>
>   compiler/typecheck/TcTyClsDecls.lhs |   26 +++++++++++++-------------
>   1 files changed, 13 insertions(+), 13 deletions(-)
>
> | -----Original Message-----
> | From: glasgow-haskell-users-bounces at haskell.org [mailto:glasgow-haskell-
> | users-bounces at haskell.org] On Behalf Of Simon Peyton-Jones
> | Sent: 05 June 2012 12:38
> | To: JP Moresmau; glasgow-haskell-users at haskell.org
> | Subject: RE: Source Location of DataCon objects in GHC 7.4.1 API
> |
> | Ah I see.  I'm fixing this.
> |
> | Simon
> |
> | | -----Original Message-----
> | | From: glasgow-haskell-users-bounces at haskell.org
> | | [mailto:glasgow-haskell- users-bounces at haskell.org] On Behalf Of JP
> | | Moresmau
> | | Sent: 01 June 2012 10:25
> | | To: glasgow-haskell-users at haskell.org
> | | Subject: Source Location of DataCon objects in GHC 7.4.1 API
> | |
> | | Hello
> | |
> | | I have a failing test in BuildWrapper when moving from GHC 7.0.4 to
> | | 7.4.1. As far I can tell, in the TypecheckedSource I get DataCon
> | | objects that have no location info, and hence I can't retrieve them by
> | | location... Which is useful in a IDE (tell me what's under my mouse
> | | cursor, tell me where else it's used).
> | |
> | | Given the simple data declaration:
> | | data DataT=MkData {name :: String}
> | |
> | | In 7.0.4 I obtain a hierarchy that ends in FunBind (on a Var called
> | | "name")/MatchGroup/Match/ConPatOut and the contained DataCon named
> | | "MkData" has a SrcSpan associated with it, and so do the Var,
> | | MatchGroup and Match.
> | | In 7.4.1 I have the same hierarchy but the DataCon tells me "no
> | | location info". The Var "name" has a location, but the MatchGroup and
> | | Match don't either.
> | |
> | | Is it a normal change? Do I need to change something in the way I load
> | | the module? Is it a regression?
> | |
> | | Thanks!
> | |
> | | --
> | | JP Moresmau
> | | http://jpmoresmau.blogspot.com/
> | |
> | | _______________________________________________
> | | Glasgow-haskell-users mailing list
> | | Glasgow-haskell-users at haskell.org
> | | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users
> |
> |
> |
> | _______________________________________________
> | Glasgow-haskell-users mailing list
> | Glasgow-haskell-users at haskell.org
> | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users
>
>
>
> _______________________________________________
> Glasgow-haskell-users mailing list
> Glasgow-haskell-users at haskell.org
> http://www.haskell.org/mailman/listinfo/glasgow-haskell-users