As far as I know it from ANSI C and some other language APIs, writing to stderr is always unbuffered, so all error message are immediately written, so mostly all messages are out before an application crashes.
Writing to stderr is rare, so the slow down for unbuffered I/O can be ignored.
Regards
Patrick
-----------------------------------
Wissenschaftliche Hilfskraft
Technische Universität Dresden
Fakultät Informatik
Institut für Technische Informatik
Lehrstuhl VLSI-Entwurfssysteme, Diagnostik und Architektur
01062 Dresden
-----Original Message-----
From: Ghdl-discuss [mailto:ghdl-discuss-boun...@gna.org] On Behalf Of Jonas
Baggett
Sent: Saturday, June 25, 2016 11:01 AM
To: Tristan Gingold <tging...@free.fr>; GHDL discuss list <ghdl-discuss@gna.org>
Subject: Re: [Ghdl-discuss] Collaboration in the GHDL project
Hi Tristan,
I have already made an --unbuffered option work. It disable buffering on
stdout, and does line buffering on stderr. Should I also completely disable
buffering on stderr ?
Currently, a C function is called to disable buffering after reading the
--unbuffered option in grt-options.adb. Does it work that way or should I call
setvbuf()/setbuf() before each write to stdout/stderr ?
What is remaining to do is to make this option work also when a file is open in
write mode in VHDL.
Jonas
Le 23. 06. 16 à 20:43, Tristan Gingold a écrit :
> On 23/06/16 20:40, Jonas Baggett wrote:
>>
>>
>> Le 23. 06. 16 à 08:01, Tristan Gingold a écrit :
>>> On 23/06/16 07:43, Jonas Baggett wrote:
>>>> Hi Tristan,
>>>>
>>>> I if understand well, what needs to be done is to add the
>>>> possibility that each write to either the console or a file can be
>>>> unbuffered, right ? Or maybe writes to the console is already
>>>> unbuffered. This feature can be done with the the C function fflush
>>>> which is already imported in grt-stdio.ads and then call it at the
>>>> end of the Put procedures in grt-astdio.adb, right ?
>>>
>>> I would simply call setvbuf() (or setbuf) on stdout/stderr and after
>>> fopen. That would be slightly more efficient.
>>>
>>> Tristan.
>>>
>> I think I will start with this issue, finally. Should the unbuffered
>> option be passed during elaboration ?
>
> No, at execution time. There are already an infrastructure to handle
> options, see grt-options.adb
>
> Tristan.
_______________________________________________
Ghdl-discuss mailing list
Ghdl-discuss@gna.org
https://mail.gna.org/listinfo/ghdl-discuss
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ Ghdl-discuss mailing list Ghdl-discuss@gna.org https://mail.gna.org/listinfo/ghdl-discuss
