CCHD@lure.latrobe.edu.au (Huw Davies - La Trobe University Computer Centre) (02/14/91)
I have a slight problem with the AIX v3 c compiler (on an RS/6000). The following code doesn't do what I want :-) Please note that this is an example of the error extracted from debugging 6K lines of c code (It took a while to find....). $ more test_struct.c #include <stdio.h> struct test { short a; int b; } c; main() { printf("a: %x\nb: %x\nSize of a: %d\n", &c.a, &c.b, (int) &c.b - (int) &c.a); } $ ./test a: 2003e460 b: 2003e464 Size of a: 4 $ I would have liked the difference in address to be 2, not 4. I know that there are efficiency reasons for aligned access, but I would like to be able to override the default. I have searched high and low all over our CD-ROM and can't find an option to the c compiler to force the slow (but nice) way. Huw Davies Computing Services e-mail: cchd@lucifer.latrobe.edu.au
bengsig@dk.oracle.com (Bjorn Engsig) (02/14/91)
Article <5053@lure.latrobe.edu.au> by CCHD@lure.latrobe.edu.au (Huw Davies - La Trobe University Computer Centre) says: | |struct test |{ | short a; | int b; |} c; | | printf("a: %x\nb: %x\nSize of a: %d\n", &c.a, &c.b, | (int) &c.b - (int) &c.a); | |a: 2003e460 |b: 2003e464 |Size of a: 4 | |I would have liked the difference in address to be 2, not 4. The ANSI C standard does actually allow the implementor to add padding between any structure member (not before the first one), so you should not make assumptions about this. I'm afraid you will need to rewrite your code to make it behave the way you want, i.e. declare the structure differently and do the packing in and out yourself. | I have searched high and low all |over our CD-ROM and can't find an option to the c compiler to force |the slow (but nice) way. Personally, I don't see what is nice about it, but I do of course not know your application. -- Bjorn Engsig, ORACLE Corporation, E-mail: bengsig@oracle.com, bengsig@oracle.nl "Stepping in others footsteps, doesn't bring you ahead"
marc@marc.watson.ibm.com (Marc Auslander) (02/14/91)
The trouble with the sort of packing you propose is that it does not provide "left to right equivalence". In other words, adding something to the end of a structure can change the mappings of fields that precede it. Consider the trouble this causes when you need to extend a structure without recompiling the world. -- Marc Auslander <marc@ibm.com>
johnp@shared.uucp (John Plattner) (02/16/91)
In article <1254@dkunix9.dk.oracle.com> bengsig@dk.oracle.com (Bjorn Engsig) writes: >Article <5053@lure.latrobe.edu.au> by CCHD@lure.latrobe.edu.au (Huw Davies - La Trobe University Computer Centre) says: >| >|struct test >|{ >| short a; >| int b; >|} c; >| >| printf("a: %x\nb: %x\nSize of a: %d\n", &c.a, &c.b, >| (int) &c.b - (int) &c.a); >| >|a: 2003e460 >|b: 2003e464 >|Size of a: 4 >| >|I would have liked the difference in address to be 2, not 4. >The ANSI C standard does actually allow the implementor to add padding >between any structure member (not before the first one), so you should not >make assumptions about this. I'm afraid you will need to rewrite your code >to make it behave the way you want, i.e. declare the structure differently >and do the packing in and out yourself. >Personally, I don't see what is nice about it, but I do of course not know >your application. If you REALLY want to bash IBM a little bit, get their COBOL (shudder shudder) compiler and create an external COBOL structure with : 01 test external. 02 a pic 9(4). 02 b pic 9(9). and try to share it with the equivalent C global structure. You will find that the COBOL compiler does NOT boundary align, and you get nice weird results. IBM's only response is "well, you'll have to manually align the COBOL structure". -- John Plattner Shared Financial Systems uunet!shared!johnp Dallas, Tx Phone: 214-458-3839 Fax: 214-458-3876 --->>> My comments are my opinions and may not be shared by Shared <<<---
bengsig@dk.oracle.com (Bjorn Engsig) (02/18/91)
Article <1991Feb15.220246.35565@shared.uucp> by johnp@shared.uucp (John Plattner) says: |If you REALLY want to bash IBM a little bit, get their COBOL (shudder shudder) | [ example ] |and try to share it with the equivalent C global structure. You will find that |the COBOL compiler does NOT boundary align, and you get nice weird results. This behaviour is not specific to IBM but to Micro Focus cobol, which is what is behind the IBM VS Cobol. I'm on thin ice here, but I actually think cobol is required to behave like that (because of data file portability); please correct me if I'm wrong. -- Bjorn Engsig, ORACLE Corporation, E-mail: bengsig@oracle.com, bengsig@oracle.nl "Stepping in others footsteps, doesn't bring you ahead"
bengsig@dk.oracle.com (Bjorn Engsig) (02/18/91)
Article <MARC.91Feb14091358@marc.watson.ibm.com> by marc@marc.watson.ibm.com (Marc Auslander) says: |Consider the trouble this causes when you need to extend |a structure without recompiling the world. Some people in the real world are used to do that on many occasions. Luckily we have make. -- Bjorn Engsig, ORACLE Corporation, E-mail: bengsig@oracle.com, bengsig@oracle.nl "Stepping in others footsteps, doesn't bring you ahead"
andreas@easix.gun.de (Andreas Baess) (02/26/91)
bengsig@dk.oracle.com (Bjorn Engsig) writes: >|If you REALLY want to bash IBM a little bit, get their COBOL (shudder shudder) >| [ example ] >|and try to share it with the equivalent C global structure. You will find that >|the COBOL compiler does NOT boundary align, and you get nice weird results. >This behaviour is not specific to IBM but to Micro Focus cobol, which >is what is behind the IBM VS Cobol. I'm on thin ice here, but I actually think >cobol is required to behave like that (because of data file portability); please >correct me if I'm wrong. This is common in all COBOL implementations I know. There once upon was a time where main storage was so incredible expensive. Every COBOL programmer knows, that alignment of data can be enforced be declaring the field with an 01 level or by adding an sync behind the declaration. -- -> andreas@easix.GUN.de {tmpmbx,mcshh,smurf,flyer}!easix -> Was man nicht begreift kann man wenigstens nicht vergessen!