DoCon as GHC test

Serge D. Mechveliani mechvel at botik.ru
Mon Oct 20 14:15:14 EDT 2003


Dear GHC developers,

I wonder: why do not you use the DoCon program as the test?

For many years, an attempt 
  to build DoCon according to  install.txt  
  and to run its test  .../demotest/T_  

reveals on average a couple of serious bugs in a new GHC version.

One does not need to study anything:
only download the archive and follow  install.txt.

Anyway, if the test function crashes, then it, probably, does not 
worth to release a public version until the bug is fixed 
-- ?
After the release is done public, the users will still find further 
bugs, but this is a regular process.

Though the test is not done specially for debugging GHC, it proves 
useful for this.
It runs first certain simple functionality, then proceeds with 
more complex. It also prints 
                             True, True ...,
and where it prints first  False  or where it crashes, it follows
then that the bug had occured before.

Generally, the situation is as follows.

* docon-2.06  is public, it is for old  ghc,  maybe, ghc-5.04.

* docon-2.07  is used privately, it is intermediate, 
              it runs on  ghc-5.04  
              and does not reveal any serious bug.

* docon-2.08  is ready to become public, it is for  ghc >= 6.0.1,
              but needs a reliable ghc release.

* As to the implemented computer algebra methods themself, 
  there is no serious progress from docon-2.06.
  But the  porting  is good, because 
  a) it is useful for the GHC debugging,
  b) DoCon has 1-2 users, now and then other people threaten to 
     use it, and this alarms me,
  c) the last GHC release is always most supported, and often 
     DoCon cannot install for new release without some changes, 
     usually, `system', cosmetic ones.

* Sometimes I drop upgrading DoCon to new GHC for an year or two,
  and this leads to extra unfixed bugs in GHC. 

* Recently I try a project of  CA + Prover based on Term Rewriting
  and so far implement half of it in Haskell.
  I tried many scientific programming tools in my life and do not 
  fear any of looking into new ones, 
  but (Glasgow) Haskell occurs hard to quit with.


With kind regards,

-----------------
Serge Mechveliani
mechvel at botik.ru








More information about the Glasgow-haskell-users mailing list