(Negative) impact of Jansi

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

(Negative) impact of Jansi

Falko Modler
Hi everyone!

I recently ran into problems when trying to get colored log output from
Quarkus tests (via Surefire): https://github.com/fusesource/jansi/issues/171

Coincidentally, we started seeing Surefire JVM fork crashes on various
Windows 10 workstations when Jansi was active (sorry, I do not have more
details at the moment).

I also found out that building Quarkus (651 modules) is 40% faster with
-Djansi.passthrough=true (using Git Bash on Windows 10 & TERM=cygwin,
and I guess also with xterm-256color).

Given these three observations, I am wondering whether it would be
better if Maven actively used -Djansi.passthrough=true for TERM variants
that are known to handle coloring on their own?
Or at least document this?

Best regards,

Falko


---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: (Negative) impact of Jansi

Hervé BOUTEMY
on crashes, I don't have any clue: details welcome

but on the performance impact of Jansi on Windows (to detect ANSI codes) when
there is much content (like with Surefire output), there is some partial work
started to improve by removing flush:
see https://github.com/fusesource/jansi/pull/150

If we have multiple people using Windows to help test and improve this, we
could probably finally merge the update

Regards,

Hervé

Le dimanche 28 juin 2020, 19:29:11 CEST Falko Modler a écrit :

> Hi everyone!
>
> I recently ran into problems when trying to get colored log output from
> Quarkus tests (via Surefire): https://github.com/fusesource/jansi/issues/171
>
> Coincidentally, we started seeing Surefire JVM fork crashes on various
> Windows 10 workstations when Jansi was active (sorry, I do not have more
> details at the moment).
>
> I also found out that building Quarkus (651 modules) is 40% faster with
> -Djansi.passthrough=true (using Git Bash on Windows 10 & TERM=cygwin,
> and I guess also with xterm-256color).
>
> Given these three observations, I am wondering whether it would be
> better if Maven actively used -Djansi.passthrough=true for TERM variants
> that are known to handle coloring on their own?
> Or at least document this?
>
> Best regards,
>
> Falko
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]





---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: (Negative) impact of Jansi

Falko Modler
> on crashes, I don't have any clue: details welcome
I see what I can do but please don't expect anything.

> see https://github.com/fusesource/jansi/pull/150
>
> If we have multiple people using Windows to help test and improve this, we
> could probably finally merge the update
I'll try to support.

But apart from that improvement, wouldn't passthrough _still_ have less
overhead?
I mean, why even have it handled by Jansi at all when the shell itself
can handle the sequences?

Cheers,
Falko

Am 29.06.2020 um 07:42 schrieb Hervé BOUTEMY:

> on crashes, I don't have any clue: details welcome
>
> but on the performance impact of Jansi on Windows (to detect ANSI codes) when
> there is much content (like with Surefire output), there is some partial work
> started to improve by removing flush:
> see https://github.com/fusesource/jansi/pull/150
>
> If we have multiple people using Windows to help test and improve this, we
> could probably finally merge the update
>
> Regards,
>
> Hervé
>
> Le dimanche 28 juin 2020, 19:29:11 CEST Falko Modler a écrit :
>> Hi everyone!
>>
>> I recently ran into problems when trying to get colored log output from
>> Quarkus tests (via Surefire): https://github.com/fusesource/jansi/issues/171
>>
>> Coincidentally, we started seeing Surefire JVM fork crashes on various
>> Windows 10 workstations when Jansi was active (sorry, I do not have more
>> details at the moment).
>>
>> I also found out that building Quarkus (651 modules) is 40% faster with
>> -Djansi.passthrough=true (using Git Bash on Windows 10 & TERM=cygwin,
>> and I guess also with xterm-256color).
>>
>> Given these three observations, I am wondering whether it would be
>> better if Maven actively used -Djansi.passthrough=true for TERM variants
>> that are known to handle coloring on their own?
>> Or at least document this?
>>
>> Best regards,
>>
>> Falko
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: [hidden email]
>> For additional commands, e-mail: [hidden email]
>
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]
>


---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: (Negative) impact of Jansi

Hervé BOUTEMY
Le lundi 29 juin 2020, 23:11:09 CEST Falko Modler a écrit :

> > on crashes, I don't have any clue: details welcome
>
> I see what I can do but please don't expect anything.
>
> > see https://github.com/fusesource/jansi/pull/150
> >
> > If we have multiple people using Windows to help test and improve this, we
> > could probably finally merge the update
>
> I'll try to support.
thank you

>
> But apart from that improvement, wouldn't passthrough _still_ have less
> overhead?
sure

> I mean, why even have it handled by Jansi at all when the shell itself
> can handle the sequences?
Windows shell can't handle ANSI escape code: that's the purpose of Jansi, and
what is represented in JAnsi front page http://fusesource.github.io/jansi/

What makes you say that a Windows shell can handle the sequences?

Regards,

Hervé

>
> Cheers,
> Falko
>
> Am 29.06.2020 um 07:42 schrieb Hervé BOUTEMY:
> > on crashes, I don't have any clue: details welcome
> >
> > but on the performance impact of Jansi on Windows (to detect ANSI codes)
> > when there is much content (like with Surefire output), there is some
> > partial work started to improve by removing flush:
> > see https://github.com/fusesource/jansi/pull/150
> >
> > If we have multiple people using Windows to help test and improve this, we
> > could probably finally merge the update
> >
> > Regards,
> >
> > Hervé
> >
> > Le dimanche 28 juin 2020, 19:29:11 CEST Falko Modler a écrit :
> >> Hi everyone!
> >>
> >> I recently ran into problems when trying to get colored log output from
> >> Quarkus tests (via Surefire):
> >> https://github.com/fusesource/jansi/issues/171
> >>
> >> Coincidentally, we started seeing Surefire JVM fork crashes on various
> >> Windows 10 workstations when Jansi was active (sorry, I do not have more
> >> details at the moment).
> >>
> >> I also found out that building Quarkus (651 modules) is 40% faster with
> >> -Djansi.passthrough=true (using Git Bash on Windows 10 & TERM=cygwin,
> >> and I guess also with xterm-256color).
> >>
> >> Given these three observations, I am wondering whether it would be
> >> better if Maven actively used -Djansi.passthrough=true for TERM variants
> >> that are known to handle coloring on their own?
> >> Or at least document this?
> >>
> >> Best regards,
> >>
> >> Falko
> >>
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: [hidden email]
> >> For additional commands, e-mail: [hidden email]
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: [hidden email]
> > For additional commands, e-mail: [hidden email]
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]





---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: (Negative) impact of Jansi

Falko Modler
> Windows shell can't handle ANSI escape code: that's the purpose of Jansi, and
> what is represented in JAnsi front page http://fusesource.github.io/jansi/
>
> What makes you say that a Windows shell can handle the sequences?

Windows itself should be able to do this with this registry hack but I
have not checked the combination of this hack together with passthrough:
https://superuser.com/a/1300251

What I meant (and written a bit vaguely) is that Git Bash on Windows
doesn't need Jansi but with this shell I encountered all these problems.

Cheers,

Falko

Am 29.06.2020 um 23:26 schrieb Hervé BOUTEMY:

> Le lundi 29 juin 2020, 23:11:09 CEST Falko Modler a écrit :
>>> on crashes, I don't have any clue: details welcome
>> I see what I can do but please don't expect anything.
>>
>>> see https://github.com/fusesource/jansi/pull/150
>>>
>>> If we have multiple people using Windows to help test and improve this, we
>>> could probably finally merge the update
>> I'll try to support.
> thank you
>
>> But apart from that improvement, wouldn't passthrough _still_ have less
>> overhead?
> sure
>
>> I mean, why even have it handled by Jansi at all when the shell itself
>> can handle the sequences?
> Windows shell can't handle ANSI escape code: that's the purpose of Jansi, and
> what is represented in JAnsi front page http://fusesource.github.io/jansi/
>
> What makes you say that a Windows shell can handle the sequences?
>
> Regards,
>
> Hervé
>
>> Cheers,
>> Falko
>>
>> Am 29.06.2020 um 07:42 schrieb Hervé BOUTEMY:
>>> on crashes, I don't have any clue: details welcome
>>>
>>> but on the performance impact of Jansi on Windows (to detect ANSI codes)
>>> when there is much content (like with Surefire output), there is some
>>> partial work started to improve by removing flush:
>>> see https://github.com/fusesource/jansi/pull/150
>>>
>>> If we have multiple people using Windows to help test and improve this, we
>>> could probably finally merge the update
>>>
>>> Regards,
>>>
>>> Hervé
>>>
>>> Le dimanche 28 juin 2020, 19:29:11 CEST Falko Modler a écrit :
>>>> Hi everyone!
>>>>
>>>> I recently ran into problems when trying to get colored log output from
>>>> Quarkus tests (via Surefire):
>>>> https://github.com/fusesource/jansi/issues/171
>>>>
>>>> Coincidentally, we started seeing Surefire JVM fork crashes on various
>>>> Windows 10 workstations when Jansi was active (sorry, I do not have more
>>>> details at the moment).
>>>>
>>>> I also found out that building Quarkus (651 modules) is 40% faster with
>>>> -Djansi.passthrough=true (using Git Bash on Windows 10 & TERM=cygwin,
>>>> and I guess also with xterm-256color).
>>>>
>>>> Given these three observations, I am wondering whether it would be
>>>> better if Maven actively used -Djansi.passthrough=true for TERM variants
>>>> that are known to handle coloring on their own?
>>>> Or at least document this?
>>>>
>>>> Best regards,
>>>>
>>>> Falko
>>>>
>>>>
>>>> ---------------------------------------------------------------------
>>>> To unsubscribe, e-mail: [hidden email]
>>>> For additional commands, e-mail: [hidden email]
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: [hidden email]
>>> For additional commands, e-mail: [hidden email]
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: [hidden email]
>> For additional commands, e-mail: [hidden email]
>
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]
>


---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: (Negative) impact of Jansi

Tibor Digana
In reply to this post by Falko Modler
>> we started seeing Surefire JVM fork crashes on various Windows 10
workstations when Jansi was active

Can you show me the log how it crashed, and maybe the dump file from
target/surefire-reports?

On Sun, Jun 28, 2020 at 7:29 PM Falko Modler <[hidden email]> wrote:

> Hi everyone!
>
> I recently ran into problems when trying to get colored log output from
> Quarkus tests (via Surefire):
> https://github.com/fusesource/jansi/issues/171
>
> Coincidentally, we started seeing Surefire JVM fork crashes on various
> Windows 10 workstations when Jansi was active (sorry, I do not have more
> details at the moment).
>
> I also found out that building Quarkus (651 modules) is 40% faster with
> -Djansi.passthrough=true (using Git Bash on Windows 10 & TERM=cygwin,
> and I guess also with xterm-256color).
>
> Given these three observations, I am wondering whether it would be
> better if Maven actively used -Djansi.passthrough=true for TERM variants
> that are known to handle coloring on their own?
> Or at least document this?
>
> Best regards,
>
> Falko
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]
>
>