 florian
|
f56aa9abe7
* set expectloc properly for x87 on x86
|
3 years ago |
 svenbarth
|
7330a0156c
* fix writing of the method and parameter names for the Variant call description
|
4 years ago |
 pierre
|
c8c6e647be
Avoid keeping optimized out node in callcleanupblock.statements or callinitblock.statements
|
4 years ago |
 nickysn
|
7ac2f9e242
* synchronized with trunk
|
4 years ago |
 yury
|
afc80a84cf
* Fixed r48650.
|
4 years ago |
|
nickysn
|
a74024e7d5
* synchronized with trunk
|
4 years ago |
|
yury
|
4245b20a32
* Fixed expectloc of local variables passed as a formal parameter. This issue is caught by cycling the compiler with -dEXTDEBUG. Yet the compiler can't cycle with EXTDEBUG due to other issues with wrong expectloc.
|
4 years ago |
|
nickysn
|
8b4aceea50
* synchronized with trunk
|
4 years ago |
 Jonas Maebe
|
5c2303a76b
* add parameter cleanup nodes to the call cleanup block (mantis #38316)
|
4 years ago |
|
nickysn
|
91e9558f5b
* synchronized with trunk
|
4 years ago |
|
florian
|
c82aae59a1
* call node_complexity correctly in order_parameters when sorting parameters
|
4 years ago |
|
svenbarth
|
1efeb5851a
* fix regression introduced with r47625 for Mantis #38122: when checking for a dereferentiation of the Self parameter subscript and vec nodes need to be kept
|
4 years ago |
|
nickysn
|
dc5b78d5f0
* synchronized with trunk
|
4 years ago |
|
svenbarth
|
69e6f3dcbb
* if a call inside a generic's code that involves generic type arguments can't be clearly determined then postpone it till specialization in the hope that the type will be clear then
|
4 years ago |
|
nickysn
|
aeafe42144
* synchronized with trunk
|
4 years ago |
|
svenbarth
|
82957ec5a3
* Fix for Mantis #38122: when a deref node is passed as a Self parameter for a type helper (which is a var parameter) we need to pass the non-derefentiated value so that the data it points to can be modified by the helper's method (this is Delphi compatible)
|
4 years ago |
|
nickysn
|
48b81da877
* synchronized with trunk
|
4 years ago |
|
florian
|
3baddff71e
+ inline subroutines with formal parameters as well
|
4 years ago |
|
nickysn
|
446d824ad6
* synchronized with trunk
|
4 years ago |
|
florian
|
ab9fa2012b
* inline more aggressively, if the ie for the initial change re-appears, it must be fixed differently
|
4 years ago |
|
nickysn
|
38c4c93cee
* synchronized with trunk
|
4 years ago |
|
florian
|
a628c1c3dd
* if an array constructor is passed to a formaldef parameter, it is passed as a set (if the conversion is possible), resolves #37796
|
4 years ago |
|
nickysn
|
d8a9bd4da0
* synchronized with trunk
|
4 years ago |
|
florian
|
b36597c76a
* better calculation of estimated stack frame size
|
4 years ago |
|
nickysn
|
7bab5b6483
* synchronized with trunk
|
5 years ago |
|
Jonas Maebe
|
453bfcd370
* emit the correct declaration for procdefs used to force a procname in LLVM
|
5 years ago |
|
nickysn
|
988a833836
* synchronized with trunk
|
5 years ago |
|
Jonas Maebe
|
d1d3d7cd08
* never create a regular temp for inline parameters that can be modified by
|
5 years ago |
|
yury
|
e5a923eb7a
* 8086: Proper implementation of: If a far procvar is called, it must be in a memory location. This fixes a hack to emulate CALL reg1:reg2.
|
5 years ago |
|
yury
|
aefc1be555
* Reverted accidentally committed files.
|
5 years ago |
