Pointer arithmetic about boolector

[tom-kacvinsky]

Should note that I tried changing 3UL to 3ULL (to get a 8 byte wide constant), but then crashed elsewhere.

[arminbiere]

Well I think there is no problem here. Actually, the 'ul' (any part of it) does not seem to be necessary. It might break if 'unsigned long' as well as 'BtorNode*' are wider than 'uintptr_t' and the latter is signed. Both conditions are not satisfied (on any system I think). To make it easier to understand we could replace it with static inline BtorNode * btor_node_real_addr (const BtorNode *node) { assert (sizeof (BtorNode*) == sizeof (sizeof (uintptr_t)); return (BtorNode *) (~ (uintptr_t) 3 & (uintptr_t) node); } Armin

…

On Tue, Oct 1, 2019 at 7:41 PM tom-kacvinsky ***@***.***> wrote: Should note that I tried changing 3UL to 3ULL (to get a 8 byte wide constant), but then crashed elsewhere. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#69?email_source=notifications&email_token=AFA2SAMSD25DP2GZD6O5V4DQMODWHA5CNFSM4I4MVA3KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEACDWUY#issuecomment-537148243>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFA2SAKYFNDUA4HW7RPZJPDQMODWHANCNFSM4I4MVA3A> .

[arminbiere]

There is one 'sizeof' too much as you might have noticed....

…

On Wed, Oct 2, 2019 at 6:31 AM Armin Biere ***@***.***> wrote: Well I think there is no problem here. Actually, the 'ul' (any part of it) does not seem to be necessary. It might break if 'unsigned long' as well as 'BtorNode*' are wider than 'uintptr_t' and the latter is signed. Both conditions are not satisfied (on any system I think). To make it easier to understand we could replace it with static inline BtorNode * btor_node_real_addr (const BtorNode *node) { assert (sizeof (BtorNode*) == sizeof (sizeof (uintptr_t)); return (BtorNode *) (~ (uintptr_t) 3 & (uintptr_t) node); } Armin On Tue, Oct 1, 2019 at 7:41 PM tom-kacvinsky ***@***.***> wrote: > Should note that I tried changing 3UL to 3ULL (to get a 8 byte wide > constant), but then crashed elsewhere. > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#69?email_source=notifications&email_token=AFA2SAMSD25DP2GZD6O5V4DQMODWHA5CNFSM4I4MVA3KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEACDWUY#issuecomment-537148243>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AFA2SAKYFNDUA4HW7RPZJPDQMODWHANCNFSM4I4MVA3A> > . >

[tom-kacvinsky]

OK< the proposal for how to handle this makes sense. To clarify, I was just pointing out the issue and what I did to work around it for the moment. I know what I did was not really a clean patch,

There is one other similar issues, again related to sizeof(long) and sizeof(void*):

#define leaf(l) ((void *) (1lu | (unsigned long) (l)))

I propose

#define leaf(l) ((void*)((uintptr_t)1 | (unitptr_t)(l)))

[aniemetz]

Well I think there is no problem here. Actually, the 'ul' (any part of it) does not seem to be necessary. It might break if 'unsigned long' as well as 'BtorNode*' are wider than 'uintptr_t' and the latter is signed. Both conditions are not satisfied (on any system I think). To make it easier to understand we could replace it with static inline BtorNode * btor_node_real_addr (const BtorNode node) { assert (sizeof (BtorNode) == sizeof (sizeof (uintptr_t)); return (BtorNode *) (~ (uintptr_t) 3 & (uintptr_t) node); }

@mpreiner and I will have a look at this and let @tom-kacvinsky know when we have a solution. @tom-kacvinsky can we ping you to test it for us?

[mpreiner]

@tom-kacvinsky can you try the branch https://github.com/Boolector/boolector/tree/cleanup-pointer-arith and see if this fixes your problems?

[tom-kacvinsky]

Will try this out today. I would have tried it this weekend but for some bizarre reason I did not get notification of the issue update until a co-worker pointed out the issue has been updated. Thanks for looking into this,

[tom-kacvinsky]

This worked, thank you very much.

[mpreiner]

This is now merged to master 66d0f7b.