Opened 11 years ago
Closed 11 years ago
#2375 closed task (fixed)
Regress check 9.3 windows builds under EDB instead of mingw
| Reported by: | robe | Owned by: | robe |
|---|---|---|---|
| Priority: | medium | Milestone: | PostGIS 2.1.0 |
| Component: | QA/buildbots | Version: | 2.1.x |
| Keywords: | mingw64 | Cc: |
Description
The 9.3 windows builds starting with PostGIS 2.1.0 I am building with gcc 4.8.0 (the 64-bit uses Structured Error Handling (SEH), and 32-bit still uses (Set Jump Long Jump) sjlj error handling (dwarf I had accidentally put in in my first go round and did not work (shall I say crashes since it can't jump VC++ dll boundaries easily).
http://stackoverflow.com/questions/15670169/what-is-difference-between-sjlj-vs-dwarf-vs-seh
Anyrate now that 9.3 w64-bit with gcc 4.8 seh regresses under EDB, I'm switching the 9.3 regress tests from mingw to EDB. Have the 9.3 w32 bit set already. Should have the 9.3 w64bit ready as well tomorrow ready for testing with upcoming 2.1.0 rc1 and 2.2.0 moving forward.
I'm still going to keep 9.2 and 9.1 compiled under the old 4.5.4 sjlj chains since I don't want to rock the boat too much. If 9.3 looks all good (which I expect it will barring I didn't screw anything up), I'll may switch the 9.2 chains in a 2.0 micro to 4.8 as well. This has to be done for both 2.0 and 2.1 if I do that to prevent corrupting a 2.0 build that has 2.1 installed as well.
