Re: .ly file partially compiles, then LP crashes

[David Wright]

On Fri 28 Jul 2017 at 15:16:03 (+0200), David Kastrup wrote:
> Bernhard Kleine <address@hidden> writes:
> 
> > Am 28.07.2017 um 00:55 schrieb Guy Stalnaker:
> >> Exited with return code -1073741819
> > This has come up with the same number IIRC repeatedly.
> 
> It's Windows' helpful way to refer to a segfault.  Storing something
> more descriptive like "Segmentation violation" for several dozen
> signal-based error messages would consume too much memory needed for
> spyware. 16kB should be enough for anybody.

I don't understand what the OS would do with these error messages.
On error, the OS returns a code¹ which is handled by the caller.
When I run a program under strace, I can see the OS generating
hundreds of errors every second and they all go unreported except
as a return code. It's up to the application to decide whether to
finally report something, and what that is.

That said, having spent years tracking down 0Cn errors in IBM Fortran,
errors like Access Violation mean next to nothing on their own because
the cause could be many levels of calls and MB of code away from the
point where the faulty address value actually triggers the error.

> It's the same reason that all of ed's (the inspiration for Edlin) error
> messages are a single question mark.

I like the 16kB. Could I just point out that the code for the very
functional EDIT program (I believe Phil Hazel wrote it—he wrote its
successor Zed) occupied 32kB of memory. This single chunk of 32kB
(reentrant) was used by 70-100 people simultaneously logged on to
Phoenix/MVS running on an IBM 370/165 containing 1MB of ferrite core
memory (later increased to 4MB). Meanwhile the system was running a
heavily used batch job service.

¹ Linux has them in include/uapi/asm-generic/errno{-base,}.h
if I'm up to date; Windows will have some equivalent header
file that google will know about.

Cheers,
David.