Re: Some serious issues with the new -O option

[Stefano Lattarini]

On 04/30/2013 11:55 AM, Tim Murphy wrote:
> I'm guessing here but I imagine the main problem comes with delaying the
> results of submakes?
>
I think so, yes.  The Distributed Make from Sun implements a similar
output synchronization feature, and doesn't have the output delay problems
I have experienced with GNU make (while it shares the issues with automatic
output colorization).

> I haven't tested to see if this is how the new feature works or not. I
> don't think it's completely necessary to keep all output from one submake
> together. so turning that off might make things more interactive,
> Per-target syncing is a valid compromise.
>
Indeed.

> Regards,
> 
> Tim

I leave my original message quoted here in tis entirety, for reference:
>
> On 30 April 2013 10:48, Stefano Lattarini <address@hidden>wrote:
> 
>> I'm sorry to say that the new -O option can interact *badly*
>> with Automake-generated parallel testsuites (at least when they
>> are being run "interactively", that is, with the output of make
>> connected to a tty).
>>
>> Let me elaborate a little (if you still have doubts or objections
>> after reading the considerations below, please do not hesitate to
>> ask for clarifications).
>>
>> For long-running testsuite with lots of tests, the use of the '-O'
>> option prevents the results from the tests already run to be displayed
>> on the screen until the *whole* testsuite has run.  IMO, this destroys
>> feedback and readability of the output.  Try building GNU coreutils and
>> running its testsuite with "make -j2 -O" on a slowish system to see how
>> this can be annoying in practice.
>>
>> Moreover, since the output of the recipes is no longer connected to
>> a terminal, automatic colorization of output no longer work: no more
>> good results in green and failures in red (unless the user forces
>> them by exporting AM_COLOR_TESTS to "always").
>>
>> In defense of the '-O' option, I must say that I see how such an option
>> can be actually useful to produce more readable logs when the output is
>> being redirected to a file (so that one doesn't care about real-time
>> feedback on the console); but for most "interactive" usages, I believe
>> its downsides definitely overweight its advantages.
>>
>> So please don't ever make that option the default; if you really really
>> want to, at least put in place some smart checks that only enable '-O'
>> when the output is *not* a tty, so that we can have the best of both
>> worlds (useful feedback for interactive runs, more readable logs for
>> batch runs).
>>
>> Thanks,
>>   Stefano