Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.compilers > #1212 > unrolled thread
| Started by | laplante.p@gmail.com |
|---|---|
| First post | 2014-04-07 22:36 -0700 |
| Last post | 2014-04-12 03:31 +0000 |
| Articles | 4 — 4 participants |
Back to article view | Back to comp.compilers
Error message with bison and flex laplante.p@gmail.com - 2014-04-07 22:36 -0700
Re: Error message with bison and flex Kaz Kylheku <kaz@kylheku.com> - 2014-04-10 15:06 +0000
Re: Error message with bison and flex Hans-Peter Diettrich <DrDiettrich1@aol.com> - 2014-04-11 03:30 +0200
Re: Error message with bison and flex glen herrmannsfeldt <gah@ugcs.caltech.edu> - 2014-04-12 03:31 +0000
| From | laplante.p@gmail.com |
|---|---|
| Date | 2014-04-07 22:36 -0700 |
| Subject | Error message with bison and flex |
| Message-ID | <14-04-005@comp.compilers> |
Sometime the text return from a syntax error is very short. I would like to output 5 lines before the error and 5 lines after the error. Is this possible without affecting the error message. If I used input to read the next 5 lins, the message won't be the same and I don't want to always accumulate the 5 lines before, it will add a lot of overhead to the lexical analyser. Tx [The way I'd do it would be to write my own YY_INPUT() that slurps in the whole file to a buffer when it starts and hands the input to the lexer a line at a time, so it knows what line it's on. If there's an error, it can then scan forward and backward in the file buffer to find the five lines each way. -John]
[toc] | [next] | [standalone]
| From | Kaz Kylheku <kaz@kylheku.com> |
|---|---|
| Date | 2014-04-10 15:06 +0000 |
| Message-ID | <14-04-007@comp.compilers> |
| In reply to | #1212 |
On 2014-04-08, laplante.p@gmail.com <laplante.p@gmail.com> wrote: > Sometime the text return from a syntax error is very short. Usually, there is little or no program text at all in a syntax error message. > I would like to output 5 lines before the error and > 5 lines after the error. I don't see the point of this. Just output errors in a way that an IDE can understand, so you can navigate through errors visually. Then you have all the lines you want before and after the error. Speaking for myself, but I suspect the majority of others, programmers do not want the error output of a compiler inflated by an order of magnitude with snippets of the source code. [People do use lex, yacc, et al for other things than compilers that run under IDEs. -John]
[toc] | [prev] | [next] | [standalone]
| From | Hans-Peter Diettrich <DrDiettrich1@aol.com> |
|---|---|
| Date | 2014-04-11 03:30 +0200 |
| Message-ID | <14-04-008@comp.compilers> |
| In reply to | #1214 |
Kaz Kylheku schrieb: > On 2014-04-08, laplante.p@gmail.com <laplante.p@gmail.com> wrote: >> Sometime the text return from a syntax error is very short. > > Usually, there is little or no program text at all in a syntax error message. > >> I would like to output 5 lines before the error and >> 5 lines after the error. > > I don't see the point of this. Me2 > Just output errors in a way that an IDE can understand, so you can navigate > through errors visually. Then you have all the lines you want before and after > the error. > [People do use lex, yacc, et al for other things than compilers that run under IDEs. -John] In these cases the "source code" display may be even less helpful. IMO every application using parsers needs its kind of an IDE, that translates the syntactical errors into something meaningful to the user. DoDi [People do use the -C5 flag to grep. -John]
[toc] | [prev] | [next] | [standalone]
| From | glen herrmannsfeldt <gah@ugcs.caltech.edu> |
|---|---|
| Date | 2014-04-12 03:31 +0000 |
| Message-ID | <14-04-009@comp.compilers> |
| In reply to | #1215 |
Hans-Peter Diettrich <DrDiettrich1@aol.com> wrote: (snip) > In these cases the "source code" display may be even less helpful. IMO > every application using parsers needs its kind of an IDE, that > translates the syntactical errors into something meaningful to the user. When I first started programming, with OS/360 compilers, it was usual for the compiler to print an annotated listing file with some type of numbering. Ones I remember printed ISN (internal statement number) instead of input file line numbers, which you could match up on the listing. Fortran G prints a $ in the suspected error column on the next line, where Fortran H gives the column number. I suppose the printed listing file has some similarity to the full screen display in an IDE. The listing file also could contain a list of variables with their address and line number in which they were used. With cards, you had to print the file and look at it. With WYLBUR, you could fetch the file back to an editing session, search through it using editor command, such as search for the word ERROR, then purge the file without printing it. I don't know if any compilers still generate listing files with line numbers and cross references. -- glen
[toc] | [prev] | [standalone]
Back to top | Article view | comp.compilers
csiph-web