December 23, 2002 scppn.exe process crash | ||||
---|---|---|---|---|
| ||||
Walter Just had another nasty - a crash rather than an ICE - which had an interesting cause. On earlier compilers, it compiled fine, but 8.31+ it was toast. It was caused by an iterator class - a helper for a TREEVIEW-wrapping sequence class - that had previously relied on operator ==() and operator !=() being free functions. In making the code work for other compilers (notably Intel, GCC & Metrowerks) they required that the operators be changed to public const member functions. However, I forgot to remove the free functions. All other compilers (and previous versions of DMC++) happily ignored the free-functions, but 8.31+ was very cross. Don't know if this is helpful, but thought I'd let you know. I would imagine you'd rather the compiler report an error (although I'm pretty sure it's not an error - it would seem like a pretty stern situation - albeit that it is a useless thing to have) than crash. Matthew |
December 24, 2002 Re: scppn.exe process crash | ||||
---|---|---|---|---|
| ||||
Posted in reply to Matthew Wilson | I need a test case that causes the crash so I can fix it. Thanks, -Walter "Matthew Wilson" <dmd@synesis.com.au> wrote in message news:au87l2$rgh$1@digitaldaemon.com... > Walter > > Just had another nasty - a crash rather than an ICE - which had an interesting cause. On earlier compilers, it compiled fine, but 8.31+ it was > toast. > > It was caused by an iterator class - a helper for a TREEVIEW-wrapping sequence class - that had previously relied on operator ==() and operator !=() being free functions. In making the code work for other compilers (notably Intel, GCC & Metrowerks) they required that the operators be changed to public const member functions. However, I forgot to remove the free functions. All other compilers (and previous versions of DMC++) happily > ignored the free-functions, but 8.31+ was very cross. > > Don't know if this is helpful, but thought I'd let you know. I would imagine > you'd rather the compiler report an error (although I'm pretty sure it's not > an error - it would seem like a pretty stern situation - albeit that it is a > useless thing to have) than crash. > > Matthew > > |
Copyright © 1999-2021 by the D Language Foundation