Re: [boost] [Boost C++ Libraries] #3644: Getting starting simplified build from source should use a log file.

Subject: Re: [boost] [Boost C++ Libraries] #3644: Getting starting simplified build from source should use a log file.
From: Vladimir Prus (vladimir_at_[hidden])
Date: 2009-11-20 04:19:57

Paul A. Bristow wrote:

>> -----Original Message-----
>> From: Boost C++ Libraries [mailto:noreply_at_[hidden]]
>> Sent: Thursday, November 19, 2009 8:44 PM
>> Cc: boost-bugs_at_[hidden]
>> Subject: Re: [Boost C++ Libraries] #3644: Getting starting simplified build from source
>> should use a log file.

BTW, looks like replying to Trac emails does not actually get the comment
added to Trac ticket.

>> 
>> #3644: Getting starting simplified build from source should use a log file.
>> --------------------------------------+-------------------------------------
>>  Reporter:  pbristow                  |       Owner:
>>      Type:  Bugs                      |      Status:  new
>> Milestone:  Boost 1.42.0              |   Component:  Building Boost
>>   Version:  Boost 1.41.0              |    Severity:  Problem
>>  Keywords:  simplified library build  |
>> --------------------------------------+-------------------------------------
>> 
>> Comment(by vladimir_prus):
>> 
>>  Would printing this information at the end of the build address your
>>  concerns?
> 
> Well yes it would, but surely it is always better to have a log file so you can check it went OK?
> 
> Or will checking %errorlevel% tell the user that?

My plan is as follows:

1. Make 'stop of first error' the default behaviour, so if you get an error, you will surely notice.
2. If the build is successfull, print the summary of where what was installed.

It *seems* that with such arrangement, you don't have to look at the log too often. 

> But checking %errorlevel% doesn't tell you what libraries you have just built.

Do you mean that the user wants to know what library have failed to build? I think with the
above behaviour, he'll know.

On the other hand, we already have a couple libraries that are built depending on configuration
(python, mpi) and few others (math long long, wide serialization) that will become such. So,
maybe printing the summary of what was built, and why not, could still be helpful (in addition
to the current output telling where what is installed).

What do you think?

- Volodya