NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/sys/sgi/11478

Text File | 1992-07-27 | 3.4 KB | 76 lines

Newsgroups: comp.sys.sgi
Path: sparky!uunet!gatech!destroyer!gumby!yale!yale.edu!ira.uka.de!chx400!bernina!hoesel
From: [e-mail] (Frans van Hoesel)
Subject: Re: cc errors
Message-ID: <[e-mail]>
Sender: [e-mail] (USENET News System)
Organization: University of Groningen, the Netherlands
References: <[e-mail]>
Date: Mon, 27 Jul 1992 20:52:20 GMT
Lines: 64

In article <[e-mail]> [e-mail] (David B.Anderson) writes:
>In article <[e-mail]> [e-mail] (Frans van Hoesel) writes:
>>I don't like the lengthy error messages from the c-compiler.
>>They fill the screen
>>with lots of dashes even if one makes only a few mistakes.
>>This behavior is *NOT* what is described in the ansi-c manual from sgi.
>>there is some page describing exactly what the format of the
>>error messages is. The format described there is not the format I get.
>>Any way to force this output format?
>>(this is still 4.0.1)
>>
>>-- frans
>
>Sorry you don't like the messages.  While I do :-), it clearly would be
>desirable to give users control over verbosity.   Can't promise when
>that will happen.   I believe it will happen, though.
>
>The dashes attempt to show exactly where the compiler thinks the error is.
>If there are lots of dashes then you have very long source lines (after
>preprocessing).    On really long lines the dashes are sometimes not
>helpful.   Tough for the compiler to know which are useful.

Having a lines that normally fits in 80 columns, the message outputs indents
it with another 6 blanks, and whenever the line crosses the border the compiler
generates 4 lines (2 lines containing the source text and 2 lines
containing the dashes), together with the message itself (1 or 2 lines)
I get on a 30 line screen only 6-7 messages. together with all the mutiple
errormessages from one line I simply mis the overview of what is going on.

The CC compiler does a much better job, and just for that reason I use the CC
compiler instead of cc, until the final stage. A bit clumsy... to compile
a program twice (CC & cc), but much more productive.

As some famous (but to me nameless) speaker one said:
a good dashboard has only one light with with a big red qustion mark on it.
If anything is wrong with the car, an experienced driver will know what is
wrong when the light flashes (well this is not a very exact quote, and he
said it propably much better, but it illustrates the difference between
CC and cc messages)
>
>The present output of *multiple* messages on a single error inflates the
>output a lot.  This was never *desirable*, but did make compiler maintenance
>easier (for reasons I won't go into).
(I guessed that, and can understand it.
It would be great if you changed that too, but it was not the
point of my request.

>Thanks for the input.
And hopefuly in the future less output (from the compiler).

If you insist on the dashes, can't you just underline it (that would save
the lines with all the dashes), and/or merge the dashpattern of multiple errors
from the same line to just one source line listing .

I like the idea of making the verbosity controlable by the user.
>
>
>Regards,
>[ David B. Anderson             [phone]             [e-mail] ]
-- frans

and as always, but most of the time not so explicit... this kind of answers
from you and the rest of the people at sgi makes me very happy with sgi.