I am pretty sure that FPC 1.0.10 would give a runtime error when attempting to read past the end of file.

E.g.

program ReadPastEOF;
var
  i: Integer;
begin
  readln(i);
  writeln(i);
  readln(i);
  writeln(i);
end.

when offered a file with just one line containing one integer, would give a runtime error. This is the behavior I would in fact expect.

At some point in the 1.9 branch, this changed (and now it bites me). Now it silently assigns 0 to i on the second read. Even with -Ci as option or {$I+}.

Can anyone comment on this?

Should I submit a bug report?

Tom

Tom Verhoeff wrote

At some point in the 1.9 branch, this changed (and now it bites me). Now it silently assigns 0 to i on the second read. Even with -Ci as option or {$I+}.

This was a bug in 2.0 which has been fixed for 2.0.1 already.

Jonas

It is still in 2.0.1 [2005/06/01] for i386, which does have a fix for another problem at the end of a file (viz. reading when the last line is not terminated by an eoln marker).

Tom

Hi kind users,

I like ask how is the status of FPC64 under linux.

About:
- reliability (can be used to build 24x365 system tcp daemons?)
- speed (how perform the executable code)
- rtl (is good or we need call directly libc for every thing?)

Does somebody has very positive experience in FPC64 to show here?

Finally, very thank you anyway.

Roberto

Roberto Della Pasqua wrote

Hi kind users, I like ask how is the status of FPC64 under linux.

I can tell you what's the developer's opinion.

About: - reliability (can be used to build 24x365 system tcp daemons?)

I think so, as long as you don't depend on seldomly used parts of the FCL.

- speed (how perform the executable code)

Reasonable.

- rtl (is good or we need call directly libc for every thing?)

The rtl works good.

We downloaded the latest snapshot from svn:

Free Pascal Compiler version 2.0.1 [2005/09/09] for i386

and it still does not give the expected runtime error. I will submit a bug report.

Tom