Messages in this thread |  | | From | (Stephen R. van den Berg) | Date | Fri, 21 Jun 1996 10:41:43 +0200 | Subject | Re: 2.0.0 crashes solved [patch] |
| |
Michael Stiller <michael@toyland.ping.de> wrote: >> srb@cuci.nl wrote: >> You have got to be kidding. The patch you just listed seems to >> be a null-operation (i.e. it transforms the code into something that is >> functionally identical before and after the patch).
>Ok, so please explain why the machine is running with this, >but not without it.
Well, three possible explanations: 1. Sheer chance (i.e. functionality hasn't changed and you bug is still waiting to happen, but simply hasn't surfaced again (yet)). 2. The change moved around the code ever so slightly, if there is some corruption problem somewhere, then maybe this movemen was enough to conceal it and prevent it from being noticed for a longer period. 3. Your GCC has a bug which causes it to produce functionally different code between the version before and after the patch.
Number 3 can be checked (and you should). Examine the assembler output. Make sure that they are functionally equivalent. If not, send in a bugreport to the gcc list.
>If this patch is really a null operation, why it is needed in 1.2.13 ?
For 1.2.13 it was *not* a null operation. The patch looked different then as well (slightly, but significant). The difference was in the code that was replaced *by* the patch, not in the code that patch replaced it with.
>2.0 ist not free of bugs, last night i ran over one bug which should >be fixed since 1.3.89 but at some rare conditions it seems that it >may bug again.
There are still many bugs in v2.0. Many of them fatal. We all know that. Nobody said it was bugfree. -- Sincerely, srb@cuci.nl Stephen R. van den Berg (AKA BuGless).
"<Clarions sounding> *No one* expects the Spanish inquisition!"
|  |