 | |
[Info-Ingres] abf problems on fortran program
comp.databases.ingres comp.databases.ingres
 |
| | Thread Tools | Display Modes |
#11
 
| |||
| |||
|
Re: abf problems on fortran program -
01-30-2009
, 07:42 AM
|
Hi Scott et al, Well done to Kristoff Picard in Tech support, who pointed out that due toa quirk in the g77 compiler the object was defined with a trailing underscore! So all I had to do was edit the procedure definition and alter the 'Symbol' field to inclde the trailing underscrore. But, g77 wasn't quite done with us...the name add_cr already has an underscore, so g77 defines a symbol with TWO trailing underscores!! A quick edit of the Symbol field to 'add_cr__' and now the blasted thing links beautifully! Yay Team! Now that's why I pay for support! Marty -----Original Message----- From: info-ingres-boun... (AT) kettleriverconsulting (DOT) com [mailto:info-ingres-boun... (AT) kettleriverconsulting (DOT) com] On Behalf Of OldSchool Sent: 29 January 2009 16:38 To: info-ing... (AT) kettleriverconsulting (DOT) com Subject: Re: [Info-Ingres] abf problems on fortran program On Jan 28, 10:31*am, "Martin Bowes" <martin.bo... (AT) ctsu (DOT) ox.ac.uk wrote: Hi Mike, Still fighting this battle? Yes. Plus I decided to raise an issue with the Corp about this as well. The abextract program is generated by abf. Guessed that bit. As for add_cr, I assume your program is generating that link need? Is this for a crypto function? It must be. But add_cr is not a crypto function, its actually taking a report and making sure there are carriage returns at the end of the lines. I'm not sure why its necessary, it might be something peculiar to VMS, but given that its part of the application I'm hesitant to try and unpick it at this point. Marty this may have been asked elsewhere, if so apologies.... so add_cr has been defined in the abf app as a procedure, with the correct return type? _______________________________________________ Info-Ingres mailing list Info-Ing... (AT) kettleriverconsulting (DOT) comhttp://www.kettleriverconsulting.com/mailman/listinfo/info-ingres- Hide quoted text - - Show quoted text - |
That's kind of the direction I was headed. Ran into similar problems
w/ Ingres 6.2 (or was it 6.4?) on DGUX many moons ago, but couldn't
recall the exact details. When the opportunity presents, you might
take a gander at the "-fno_underscoring" option
discussed in the g77 man page. I'm not sure if that would help or
hurt in your situation.....or if sleeping dogs should be let lie.
as I recall, similar problems exist in a *lot* of mixed language
calling conventions.
Best wishes....
#12
| |||
| |||
|
|
Hi Scott et al, Well done to Kristoff Picard in Tech support, who pointed out that due to a quirk in the g77 compiler the object was defined with a trailing underscore! So all I had to do was edit the procedure definition and alter the 'Symbol' field to inclde the trailing underscrore. But, g77 wasn't quite done with us...the name add_cr already has an underscore, so g77 defines a symbol with TWO trailing underscores!! A quick edit of the Symbol field to 'add_cr__' and now the blasted thing links beautifully! Yay Team! Now that's why I pay for support! Marty -----Original Message----- From: info-ingres-boun... (AT) kettleriverconsulting (DOT) com [mailto:info-ingres-boun... (AT) kettleriverconsulting (DOT) com] On Behalf Of OldSchool Sent: 29 January 2009 16:38 To: info-ing... (AT) kettleriverconsulting (DOT) com Subject: Re: [Info-Ingres] abf problems on fortran program On Jan 28, 10:31*am, "Martin Bowes" <martin.bo... (AT) ctsu (DOT) ox.ac.uk wrote: Hi Mike, Still fighting this battle? Yes. Plus I decided to raise an issue with the Corp about this as well. The abextract program is generated by abf. Guessed that bit. As for add_cr, I assume your program is generating that link need? Is this for a crypto function? It must be. But add_cr is not a crypto function, its actually taking a report and making sure there are carriage returns at the end of the lines. I'm not sure why its necessary, it might be something peculiar to VMS, but given that its part of the application I'm hesitant to try and unpick it at this point. Marty this may have been asked elsewhere, if so apologies.... so add_cr has been defined in the abf app as a procedure, with the correct return type? _______________________________________________ Info-Ingres mailing list Info-Ing... (AT) kettleriverconsulting (DOT) comhttp://www.kettleriverconsulting.com/mailman/listinfo/info-ingres- Hide quoted text - - Show quoted text - |
That's kind of the direction I was headed. Ran into similar problems
w/ Ingres 6.2 (or was it 6.4?) on DGUX many moons ago, but couldn't
recall the exact details. When the opportunity presents, you might
take a gander at the "-fno_underscoring" option
discussed in the g77 man page. I'm not sure if that would help or
hurt in your situation.....or if sleeping dogs should be let lie.
as I recall, similar problems exist in a *lot* of mixed language
calling conventions.
Best wishes....
_______________________________________________
Info-Ingres mailing list
Info-Ingres (AT) kettleriverconsulting (DOT) com
http://www.kettleriverconsulting.com...fo/info-ingres
#13
| |||
| |||
|
#14
| |||
| |||
|
|
The GNU people warn to use -fno_underscoring: "Use of -fno- underscoring is not recommended unless ....." (see the man pages) However, I tested it anyway by changing utcom.def. When you do this you run immediateley into a new problem, missing several functions in libingresetc, because of a missing underscore. (they are defined with that underscore ....) Cheers Kristoff |
#15
| |||
| |||
|
|
On Jan 30, 10:12*am, Kristoff <kristoff.pic... (AT) ingres (DOT) com> wrote: The GNU people warn to use -fno_underscoring: "Use of -fno- underscoring is not recommended unless ....." (see the man pages) However, I tested it anyway by changing utcom.def. When you do this you run immediateley into a new problem, missing several functions in libingresetc, because of a missing underscore. (they are defined with that underscore ....) Cheers Kristoff i kinda suspected as much....thx |
Elaborating more on the IIseterr/iiseterr naming conflict by doing
some research on the symbol tables inside archive library libingres.a
The following entry is from the library implementation that will
compile with 'iiseterr'
libingres.a:iiuflibq.o:00000cf0 T iiseterr_
libingres.a:iiuflbqL.o:00000cf0 T iiseterr
libingres.a:iiaclibq.o:000008a0 T IIpseterr
we see that there are three implementations of iiseterr each one in a
different library, thus any of the three can be used in the source
code and will compile fine
In the offending implementations where the undefined reference error
occurred there is no 'iiseterr' entry (as in the manual) for any
library contained in libingres.a; however there is an IIseterr entry
and that is why when capitalizing both i's the symbol can be found and
hence compile.
(it would compile with 'iiseterr_' as well)
libingres.a:iiseterr.o:00000000 T IIseterr
libingres.a:iiuflibq.o:00000a5f T iiseterr_
(a refresher on the issue ):
http://groups.google.com/group/comp....ec53592a5fc7fb
)
|
« Previous Thread
|
Next Thread »
| Thread Tools | |
| Display Modes | |
Linear Mode
| |
Powered by vBulletin Version 3.5.3
Copyright ©2000 - 2012, Jelsoft Enterprises Ltd.
Copyright ©2000 - 2012, Jelsoft Enterprises Ltd.