Re: clang and gcc7

[mf2k]

Thanks Ken. I somehow missed that ticket. 


> On Sep 6, 2017, at 10:52 AM, Ken Cunningham  
> wrote:
> 
> see also 
> 
> On 2017-09-06, at 9:44 AM,   wrote:
> 
>> Is there a reason that clang-4.0 +libstdcxx (the default) requires gcc6? I 
>> am switching to gcc7 and would like to not depend gcc6 anymore. The same is 
>> true for clang-5.0. 
>> 
>> 
>> Cheers!
>> Frank
>> 
> 

Re: clang and gcc7

[Ken Cunningham]

see also 

On 2017-09-06, at 9:44 AM,   wrote:

> Is there a reason that clang-4.0 +libstdcxx (the default) requires gcc6? I am 
> switching to gcc7 and would like to not depend gcc6 anymore. The same is true 
> for clang-5.0. 
> 
> 
> Cheers!
> Frank
> 

clang and gcc7

[mf2k]

Is there a reason that clang-4.0 +libstdcxx (the default) requires gcc6? I am 
switching to gcc7 and would like to not depend gcc6 anymore. The same is true 
for clang-5.0. 


Cheers!
Frank

Re: crash with python3.6.2 and readline

[Ryan Schmidt]

We should keep the discussion on the mailing list. (Use Reply All when you 
reply.)

On Sep 6, 2017, at 06:35, David Epstein wrote:

> On 06/09/2017 08:58, Ryan Schmidt wrote:
>> On Sep 5, 2017, at 18:29, David Epstein wrote:
>> 
>> 
>>> What is the procedure for adding to bug report #54728.
>>> 
>> Log in to our Trac with your GitHub account. This will reveal the new 
>> comment field at the bottom of the ticket which you can type into and then 
>> submit.
>> 
> I have a GitHub account, but I don't think I have ever used it. When I logged 
> in to https://github.com I got
> Learn Git and GitHub without any code!
> 
> Using the Hello World guide, you’ll create a repository, start a branch, 
> write comments, and open a pull request.
> 
> I could click on "Read the Guide" or "Start a Project". But the second wasn't 
> relevant and the first sounded as though it would take more time than I have. 
> Also I wasn't sure whether reading the guide would enable me to comment on 
> the relevant ticket.
> ==
> All I now want to add to Ticket #54728 is this:
> 
> Action: installed python36 with no variant. Result: readline misbehaves in 
> python interpreter.
> Next Action: installed python36+readline.  Result: port notes python36 says 
> "install py36-readline"
> Next Action: installed py36-readline.  Result: python interpreter 
> crashes on receiving Carriage Return
> Next Action: deactivated py36-readline. Result: readline in python 
> interpreter appears to work perfectly
> Next Action: activated py36-readline. Result: python interpreter 
> crashes on receiving Carriage Return
> Next Action: deactivated py36-readline. Result: readline appears to work 
> perfectly
> 
> Action necessary from maintainer: please edit the note from "port install 
> python36" to improve the advice.
> ===
> Please give me more details about getting access to github's macports/trac, 
> or else please include the above lines of comment.

All you need to do is click the "GitHub Login" link at the top of any MacPorts 
Trac page, or visit: https://trac.macports.org/login

Then you can add comments to tickets or file new tickets.


> It's not clear that my experience has much to do with this ticket, but there 
> could be a common origin. Possibly a new ticket should be opened, but I don't 
> want to do that, as I understand too little of what is going on.

I think we already know that using py36-readline causes a crash, and that 
upstream may have already fixed the problem and just needs to release a new 
version.

https://trac.macports.org/ticket/53360

Incidentally, it seems that py36-gnureadline replaces py36-readline, and the 
question of removing the entire py-readline port seems to have stalled:

https://trac.macports.org/ticket/44206

But the current version of py36-gnureadline has the same problem.

https://trac.macports.org/ticket/53176

I don't think we need any additional tickets describing the same problem.

Re: crash with python3.6.2 and readline

[Ryan Schmidt]

On Sep 6, 2017, at 02:41, David Epstein wrote:

> I don't suppose that the output from a port install command is saved 
> automatically somewhere? It seems like it would often be useful to have this 
> output saved.

Nope, not saved anywhere.

> Last night I posted (with the same subject line as this email) a problem 
> about python3.6 and readline.
> This morning I seem to have fixed the problem by deactivating py36-readline 
> (but I have only done cursory tests).
> I have just activated py36-readme, and python3.6 again crashes immediately. 
> Subsequent deactivation restores sanity.
> Conclusion: Although I cannot be sure that my current version python36 
> @3.6.2_0+readline (active) solves all my readline problems, I can be sure 
> that it was BAD to install py36-readline on top of this.
> 
> I would like to check that the installation of python36+readline really did 
> advise me to install py36-readline, or did I just misread or imagine the 
> advice?
> 
> Can I check what port install told me, without having explicitly asked the 
> unix shell to save the output?

You can re-display any port's notes, e.g.:

$ port notes python36
python36 has the following notes:
  To make this the default Python or Python 3 (i.e., the version run
  by the 'python' or 'python3' commands), run one or both of:
  
  sudo port select --set python python36
  sudo port select --set python3 python36
  
  ##
  # IF YOU ARE USING PYTHON FROM THE TERMINAL, PLEASE INSTALL:
  #   py36-readline
  # TO AVOID A LIBEDIT / PYTHON INTERACTION ISSUE.
  # REF: https://trac.macports.org/ticket/48807
  ##

Re: crash with python3.6.2 and readline

[Ryan Schmidt]

On Sep 5, 2017, at 18:29, David Epstein wrote:

> What is the procedure for adding to bug report #54728.

Log in to our Trac with your GitHub account. This will reveal the new comment 
field at the bottom of the ticket which you can type into and then submit.


> Here is a summary of what happened on my machine:
> 
> 1. Installed python3.6.2 without specifying a variant.
> 2. This version is stable (that is, no crash) but interaction with readline 
> is faulty.
> 3. Installed python3.6.2 +readline.
> 4. Installed py36-readline as suggested by the installation script for the 
> previous step.
> 5. the new version of python, in interactive mode, crashes when I type 
> carriage return, with message like
>  
> Python(21762,0x7fffb987c3c0) malloc: *** error for object 0x10d930640: 
> pointer being freed was not allocated.
> 
> 6. Interaction of readline with unix shell appears normal after call to 
> python terminates (both variants).
> 
> I have some information (file attached bu1.txt). I also attach the crash 
> report produced by Macos (bug2.txt).
> I haven't yet collected any evidence for Point 2 above, but the fault could 
> perhaps be demonstrated using script.
> 
> What additional info would be helpful?
> 
> David Epstein
> 


Uninstall py36-readline. Yes, python36 tells you to install it, but don't, 
because as you see it currently causes crashes.

https://trac.macports.org/ticket/53360

Re: crash with python3.6.2 and readline

[David Epstein]

I don't suppose that the output from a port install command is saved 
automatically somewhere? It seems like it would often be useful to have 
this output saved.


Last night I posted (with the same subject line as this email) a problem 
about python3.6 and readline.
This morning I seem to have fixed the problem by deactivating 
py36-readline (but I have only done cursory tests).
I have just activated py36-readme, and python3.6 again crashes 
immediately. Subsequent deactivation restores sanity.
Conclusion: Although I cannot be sure that my current version python36 
@3.6.2_0+readline (active) solves all my readline problems, I can be 
sure that it was BAD to install py36-readline on top of this.


I would like to check that the installation of python36+readline really 
did advise me to install py36-readline, or did I just misread or imagine 
the advice?


Can I check what port install told me, without having explicitly asked 
the unix shell to save the output?


thanks
David Epstein