| Ticket UUID: | 2819200 | |||
| Title: | denormal failures on MIPS | |||
| Type: | Bug | Version: | obsolete: 8.5.7 | |
| Submitter: | dgp | Created on: | 2009-07-09 19:45:15 | |
| Subsystem: | 48. Number Handling | Assigned To: | dgp | |
| Priority: | 5 Medium | Severity: | ||
| Status: | Closed | Last Modified: | 2009-07-17 04:22:37 | |
| Resolution: | Fixed | Closed By: | dgp | |
| Closed on: | 2009-07-16 21:22:37 | |||
| Description: |
According to the test suite, Tcl expects gradual underflow in its floating point calculations and in the operations that underly parsing and string rep generation for doubles. The MIPS FPU in default setting flushes all denormals to zero, causing many test suite failures. The set_fpc_csr() routine declared in <sys/fpu.h> (on IRIX systems only?) lets us control this setting. Attached patch revises Tcl's initialization to establish the setting it needs. POTENTIAL INCOMPAT with embedding programs that need the opposite setting (do they exist?). | |||
| User Comments: |
dgp added on 2009-07-17 04:22:37:
allow_comments - 1 fixed for 8.5.8/8.6b2 dgp added on 2009-07-10 02:45:15: File Added - 334319: mipsdenorm.patch | |||
Attachments:
- mipsdenorm.patch [download] added by dgp on 2009-07-10 02:45:15. [details]