Remove warning in tcg stub (tcg_out_reloc)
PATCH 1/1 Remove warning in tcg stub (tcg_out_reloc)
On S390 we don't have a real TCG implementation but use a stub instead. This stub obviously doesn't call any of the TCG helper functions that are usually used by the other TCG targets.
If such a helper function is static though, we end up getting a warning that turns into an error thanks to Werror. So to successfully compile qemu, we need to get rid of it. In this case it's tcg_out_reloc.
To do that, I figured the easiest way is to call tcg_out_reloc with dumb parameters after an abort call. That way the code in question never gets executed, but we got rid of the warning.
--- tcg/s390/tcg-target.c | 6 ++++++ 1 files changed, 6 insertions(+), 0 deletions(-)
tcg/s390/tcg-target.c +6 -0
@@ -35,6 +35,12 @@ static void patch_reloc(uint8_t *code_ptr, int type, tcg_target_long value, tcg_target_long addend) {tcg_abort (); + +/* +* XXX We need to call this function at least once, since it's static. +* Please remove that code when implementing real s390x tcg support. +*/ +tcg_out_reloc (NULL ,NULL ,0 ,0 ,0 ); }static inline int tcg_target_get_call_iarg_regs_count (int flags)
Signature · 1 line
1.6.0.2
Am 11.05.2010 18:23, schrieb Alexander Graf:
Quoted 24 lines
On S390 we don't have a real TCG implementation but use a stub instead. This stub obviously doesn't call any of the TCG helper functions that are usually used by the other TCG targets.
If such a helper function is static though, we end up getting a warning that turns into an error thanks to Werror. So to successfully compile qemu, we need to get rid of it. In this case it's
tcg_out_reloc.To do that, I figured the easiest way is to call
tcg_out_relocwith dumb parameters after an abort call. That way the code in question never gets executed, but we got rid of the warning.Signed-off-by Alexander Graf <[email protected]>---
tcg/s390/tcg-target.c| 6 ++++++1 files changed, 6 insertions(+), 0 deletions(-)tcg/s390/tcg-target.c +0
-0@@ -35,6 +35,12 @@ static void patch_reloc(uint8_t *code_ptr, int type, tcg_target_long value, tcg_target_long addend) {
tcg/s390/tcg-target.c static void patch_reloc(uint8_t *code_ptr, int type, · line 46
Quoted 10 lines
tcg/s390/tcg-target.c +6
-0tcg_abort (); + +/* +* XXX We need to call this function at least once, since it's static. +* Please remove that code when implementing real s390x tcg support. +*/ +tcg_out_reloc (NULL ,NULL ,0 ,0 ,0 ); }static inline int tcg_target_get_call_iarg_regs_count (int flags)
Won't you get another warning about unreachable code because tcg_abort never returns?
What about condition compilation for tcg_out_reloc (don't compile it for hosts which don't need it)?
Regards,
Stefan
Quoted 2 lines · Stefan Weil
Won't you get another warning about unreachable code because
tcg_abortnever returns?
We don't enable that warning.
Quoted 2 lines
What about condition compilation for
tcg_out_reloc(don't compile it for hosts which don't need it)?
All hosts should need it. S390 doesn't only because it's all stubbed out.
Frankly, it shouldn't take much effort to get a working TCG for s390x...
r~
Am 11.05.2010 um 19:26 schrieb Richard Henderson <[email protected]>:
Quoted 6 lines · Stefan Weil
Quoted 2 lines
Won't you get another warning about unreachable code because
tcg_abortnever returns?We don't enable that warning.
Quoted 8 lines
Quoted 2 lines
What about condition compilation for
tcg_out_reloc(don't compile it for hosts which don't need it)?All hosts should need it. S390 doesn't only because it's all stubbed out.
Frankly, it shouldn't take much effort to get a working TCG for s390x...
Yeah, I have submitted a patch months ago, but Aurelien didn't get around to review it yet :(.
Alex
Quoted 1 line
Quoted 12 lines · Alexander Graf
Am 11.05.2010 um 19:26 schrieb Richard Henderson <[email protected]>:
Quoted 9 lines · Stefan Weil
Quoted 2 lines
Won't you get another warning about unreachable code because
tcg_abortnever returns?We don't enable that warning.
Quoted 2 lines
What about condition compilation for
tcg_out_reloc(don't compile it for hosts which don't need it)?
Quoted 9 lines
Quoted 5 lines
All hosts should need it. S390 doesn't only because it's all stubbed out.
Frankly, it shouldn't take much effort to get a working TCG for s390x...
Yeah, I have submitted a patch months ago, but Aurelien didn't get around to review it yet :(.
This is indeed the best solution. It's still on my TODO list, just that I have limited time and it's always easier to choose to ignore one big patch and got complains from a single person, than ignoring a lot of small patches and getting a lot of complains...
If only people were reviewing other people patches a bit more...
Signature · 2 lines
Aurelien Jarno GPG: 1024D/F1BCDB73 [email protected] http://www.aurel32.net
Quoted 19 lines · Aurelien Jarno
Quoted 18 lines · Alexander Graf
Am 11.05.2010 um 19:26 schrieb Richard Henderson <[email protected]>:
Quoted 14 lines · Stefan Weil
Quoted 2 lines
Won't you get another warning about unreachable code because
tcg_abortnever returns?We don't enable that warning.
Quoted 2 lines
What about condition compilation for
tcg_out_reloc(don't compile it for hosts which don't need it)?All hosts should need it. S390 doesn't only because it's all stubbed out.
Frankly, it shouldn't take much effort to get a working TCG for s390x...
Quoted 8 lines
Quoted 3 lines
Yeah, I have submitted a patch months ago, but Aurelien didn't get around to review it yet :(.
This is indeed the best solution. It's still on my TODO list, just that I have limited time and it's always easier to choose to ignore one big patch and got complains from a single person, than ignoring a lot of small patches and getting a lot of complains...
I can split up the patch if you like :).
Quoted 1 line
If only people were reviewing other people patches a bit more...
Maybe Richard could also help to review it? If so, I can certainly rebase it to current git and send it off again.
Alex
Quoted 23 lines · Alexander Graf
On S390 we don't have a real TCG implementation but use a stub instead. This stub obviously doesn't call any of the TCG helper functions that are usually used by the other TCG targets.
If such a helper function is static though, we end up getting a warning that turns into an error thanks to Werror. So to successfully compile qemu, we need to get rid of it. In this case it's
tcg_out_reloc.To do that, I figured the easiest way is to call
tcg_out_relocwith dumb parameters after an abort call. That way the code in question never gets executed, but we got rid of the warning.Signed-off-by Alexander Graf <[email protected]>--- tcg/s390/tcg-target.c | 6 ++++++ 1 files changed, 6 insertions(+), 0 deletions(-)tcg/s390/tcg-target.c +0
-0@@ -35,6 +35,12 @@ static void patch_reloc(uint8_t *code_ptr, int type, tcg_target_long value, tcg_target_long addend)
tcg/s390/tcg-target.c static void patch_reloc(uint8_t *code_ptr, int type, · line 45
Quoted 10 lines
tcg/s390/tcg-target.c +6
-0{tcg_abort (); + +/* +* XXX We need to call this function at least once, since it's static. +* Please remove that code when implementing real s390x tcg support. +*/ +tcg_out_reloc (NULL ,NULL ,0 ,0 ,0 ); }
What about declaring tcg_out_reloc static inline?
Signature · 2 lines
Aurelien Jarno GPG: 1024D/F1BCDB73 [email protected] http://www.aurel32.net
Quoted 5 lines · Aurelien Jarno
Quoted 3 lines
+
tcg_out_reloc(NULL, NULL, 0, 0, 0);
}What about declaring
tcg_out_relocstatic inline?
I think we're not far away from a mergable s390 port. I think the smallest local change is best in the interim.
r~