harendra (Harendra Kumar)
User

Projects

User does not belong to any projects.

User Details

User Since
Mar 21 2016, 2:34 AM (173 w, 1 d)

Recent Activity

Sep 11 2017

harendra added inline comments to D3945: Clarify seq documentation.
Sep 11 2017, 10:23 AM

Mar 27 2017

harendra accepted D3388: base: Check for path separators chars in openTempFile' template string.

LGTM. I usually prefer to fix the more than 80 char warnings, but that's just a nitpick.

Mar 27 2017, 1:07 PM

Mar 10 2017

harendra added a comment to D2940: Deprecate direct arg passing to ghc from runghc.

I was thinking of implementing escaping but there is one more problem here. What I had mentioned earlier was only Unix/Bash kind of escaping. Even though might work across Unix shells, it may not work for Windows, the windows command line might have different conventions for escaping. For example, see this http://stackoverflow.com/questions/7760545/escape-double-quotes-in-parameter . It may be better idea if we have some working, tested and maintained code rather than researching and then implementing it from scratch, miss some cases and then keep fixing, maintaining etc. Maybe a good idea to check with someone in the cabal team who knows that code, though it may or may not be easy to pluck out that code. Overall this seems to be a pretty low return on investment fix and I have been too busy at work, so I am inclined to drop this at this point rather than pursuing further.

Mar 10 2017, 12:10 PM

Feb 8 2017

harendra added a comment to D2940: Deprecate direct arg passing to ghc from runghc.

Thanks @bgamari for quick response and for writing the code snippet too! However, my main point was whether this solution (i.e. --ghc-options) is an acceptable solution or we need to seek more discussion on the whole issue. From your response and implementation it seems to be acceptable to you at least? If so, I can happily implement it.

Feb 8 2017, 4:07 PM
harendra added a comment to D2940: Deprecate direct arg passing to ghc from runghc.

Not sure what to do with this change. The trac discussion suggested to start a GHC proposal process, but I have very little time on my hands to be able to drive that. I can at most try to figure out a way to implement --ghc-options as suggested and discussed above. If that is an acceptable option I will proceed to make the changes otherwise I will have to just drop this unless someone provides a way forward.

Feb 8 2017, 2:57 PM

Jan 14 2017

harendra updated the diff for D2940: Deprecate direct arg passing to ghc from runghc.
  • Fix doc formatting, add trac issue reference
Jan 14 2017, 4:42 AM

Jan 8 2017

harendra added a comment to D2940: Deprecate direct arg passing to ghc from runghc.

Is it not more usual to have a --ghc-options flag which passes a whole string to ghc rather than specifying each atom individually?

I had thought about it and I took the same approach as taken by gcc -Xlinker option which does exactly the same thing for passing arguments to the linker. I _think_ the problem with multiple options as one argument is that we will have to tokenize them ourselves using space as separator. Which means, if we really want to pass space separated string as a single option (e.g. a file name) then it will have to be escaped and then we will have to handle the escape as well and the escape of escape. This also means that the user will have to know about/learn our quoting/escaping mechanism which may or may not be compatible with the shell.

To make it more convenient I was thinking of providing a shorthand for --ghc-arg, say -g and used like -g -Wall or -g=-Wall or both ways. Need opinions on that as well as the option name.

I'm not sure what you mean here. Most Haskell programs which invoke GHC have an option --ghc-options which you use like
--ghc-options="-Wall -fno-state-hack -ddump-simpl". What is wrong with that whole string being appended to the invocation of GHC?

Jan 8 2017, 12:44 PM
harendra added a comment to D2940: Deprecate direct arg passing to ghc from runghc.

Couple more disclosures:

Jan 8 2017, 7:31 AM
harendra added a comment to D2940: Deprecate direct arg passing to ghc from runghc.

Is it not more usual to have a --ghc-options flag which passes a whole string to ghc rather than specifying each atom individually?

Jan 8 2017, 7:23 AM

Jan 7 2017

harendra added a reviewer for D2940: Deprecate direct arg passing to ghc from runghc: bgamari.
Jan 7 2017, 10:05 PM

Aug 31 2016

harendra added a comment to D2463: Fix and complete runghc documentation.

No problem. Thanks for the tips. Now I know what to do next time.

Aug 31 2016, 12:55 PM

Aug 29 2016

harendra added a comment to D2463: Fix and complete runghc documentation.

Thanks! I agree with your suggestion. Now that you have accepted it, I am not sure if I should fix and upload again. I have a couple of process/phabricator related questions:

Aug 29 2016, 3:00 PM

Aug 28 2016

thomie awarded D2464: Clarify pkg selection when multiple versions are available a Like token.
Aug 28 2016, 5:35 PM
harendra added a reviewer for D2463: Fix and complete runghc documentation: thomie.
Aug 28 2016, 8:22 AM

Aug 25 2016

harendra added a comment to D2470: Fix doc build inconsistency.

I agree, it can be fixed that way. But what I was worried about was that this will be a point fix for the doc targets only and not a general mechanism for other such problems. And make is hard to debug so we should venture more time only if we get a significant return.

Aug 25 2016, 10:41 AM

Aug 24 2016

harendra added a comment to D2464: Clarify pkg selection when multiple versions are available.

I uploaded another commit. I think I addressed all the comments. Please go through it again and let me know if there is anything which is inaccurate or needs to be reworked.

Aug 24 2016, 6:29 AM
harendra updated the diff for D2464: Clarify pkg selection when multiple versions are available.
  • Packages doc: address review comments by ezyang
Aug 24 2016, 6:23 AM
harendra added inline comments to D2464: Clarify pkg selection when multiple versions are available.
Aug 24 2016, 12:44 AM

Aug 23 2016

harendra added a comment to D2470: Fix doc build inconsistency.

That's a general and a bit more difficult problem to solve. It's not just doc targets, it's true for any targets that's been disabled, that's the way build is system is designed. To be able to do that we will have to know which targets potentially exist but have been disabled. Also for each target we will perhaps want to print specific help info as well about how to enable it. It will require a lot more work.

Aug 23 2016, 11:16 PM
harendra retitled D2470: Fix doc build inconsistency from to Fix doc build inconsistency.
Aug 23 2016, 1:53 PM
harendra added inline comments to D2464: Clarify pkg selection when multiple versions are available.
Aug 23 2016, 9:13 AM
harendra added inline comments to D2464: Clarify pkg selection when multiple versions are available.
Aug 23 2016, 8:27 AM
harendra added a comment to D2464: Clarify pkg selection when multiple versions are available.

Please pardon my ignorance. This was the first time I changed anything in GHC code base or used arc/phabricator. I hoped it will show separate commits like github. But I realized only after uploading that it shows only the combined diff. I have a related question, when I make more changes and upload to the same diff, what happens to the previous review/comment history, is it retained by the tool? Does it get affected in any way?

Aug 23 2016, 2:33 AM

Aug 22 2016

harendra updated the diff for D2464: Clarify pkg selection when multiple versions are available.

Adding a few more commits related to package databases' documentation

  • Update the documentation on package databases
  • Move "package environments" section
  • Retain a space between a flag and its argument
Aug 22 2016, 2:04 PM
harendra retitled D2464: Clarify pkg selection when multiple versions are available from to Clarify pkg selection when multiple versions are available.
Aug 22 2016, 9:45 AM
harendra retitled D2463: Fix and complete runghc documentation from to Fix and complete runghc documentation.
Aug 22 2016, 8:12 AM