View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0000848 | 1003.1(2013)/Issue7+TC1 | Shell and Utilities | public | 2014-06-19 15:17 | 2019-06-10 08:54 |
| Reporter | rhansen | Assigned To | |||
| Priority | normal | Severity | Objection | Type | Error |
| Status | Closed | Resolution | Accepted As Marked | ||
| Name | Richard Hansen | ||||
| Organization | BBN | ||||
| User Reference | |||||
| Section | c99, make | ||||
| Page Number | 2512, 2945 | ||||
| Line Number | 80970-80975, 97465 | ||||
| Interp Status | Approved | ||||
| Final Accepted Text | 0000848:0002269 | ||||
| Summary | 0000848: make defaults CFLAGS to -O, but c99 requires a level | ||||
| Description | The default value of CFLAGS in make is '-O', but the c99 '-O' option requires an optlevel option-arguement. see http://thread.gmane.org/gmane.comp.standards.posix.austin.general/9508 | ||||
| Desired Action | to be added in a bug note (allow a missing level? change the default? if change the default, what should the default be?) | ||||
| Tags | tc2-2008 | ||||
|
|
Interpretation response ------------------------ The standard states that by default make shall achieve results that are the same as CFLAGS=-O, and conforming implementations must conform to this. However, concerns have been raised about this which are being referred to the sponsor. Rationale: ------------- The specification for the c99 utility does not support -O without an option-argument. In addition, if a subsequent argument is provided, the argument will be (erroneously) interpreted as an optimization level. Notes to the Editor (not part of this interpretation): ------------------------------------------------------- On page 2945 line 97465 change: CFLAGS=-O to: CFLAGS=-O 1 On page 2949 line 97621 change: c99 −c −O $*.c to: c99 −c −O 1 $*.c On page 2949 line 97624 change: c99 −c −O $< to: c99 −c −O 1 $< |
|
|
Two thoughts: 1. Currently, a conforming application could override CC to specify a custom compiler and expect that CFLAGS should still be set to just "-O" (or not set at all). Alternative "-O optlevel" flags might not have the same semantics. E.g., "gcc" doesn't support "-O 1". 2. Also, since "-O 1" doesn't have any specified semantics for the c99 utility, so why specify "-O 1" at all? Perhaps instead simply specify that if CFLAGS is set by the system, it will be a "-O" option suitable for use with the system provided c99 utility? |
|
|
Historically, C-Compilers did not have an optimization level argument to the -O option, so I would guess that all existing compilers support -O without argument. The SunPro compiler even does not support -O space <level>: LC_ALL=C cc -O 2 t.c "t.c", line 2: warning: old-style declaration or incorrect type for: main "t.c", line 3: warning: implicit function declaration: printf ld: fatal: file 2: open failed: No such file or directory ld: fatal: file processing errors. No output written to a.out LC_ALL=C c89 -O 1 t.c c89: Warning: option -1 passed to ld usage: ld [-6:abc:d:e:f:h:il:mo:p:rstu:z:B:CD:F:GI:L:M:N:P:Q:R:S:VW:Y:?] file(s) [-64] enforce a 64-bit link-edit so the current standard requirements seems to be wrong. Interestingly, c99 behaves as the standard requires: LC_ALL=C c99 -O t.c c99: Warning: illegal option -Ot.c usage: c99 [ options] files. Use 'c99 -flags' for details so it seems that a bug in the standard created a hard to solve situation. |
|
|
The semantics of c99 -O 1 are not completely unspecified; c99 is required to set the level of code optimization to level 1. It is only the nature of level 1 optimization that is unspecified. I assume this is to allow not just for differing types of optimization but also for the possibility that some compilers might not support optimization at all, in which case -O 1 would be the same as -O 0. It is also worth noting that the macro defaults for the make utility are not literal requirements. Line 97447 says "The default rules for make shall achieve results that are the same as if the following were used." Some implementations seem to have taken advantage of this to have an empty CFLAGS by default (which is fine provided c99 without -O behaves the same as c99 -O 1). So I don't see any problem with making the change Richard proposed in 0000848:0002269. |
|
|
On Solaris, the situation is that the standard compiler cc does not accept an otimization level separated by a space and c99 does not work if the space is omitted. make only causes no problems as it does not define optimization to be turned on by default. |
|
|
Interpretation proposed June 27 2014 |
|
|
Interpretation approved 5 August 2014 |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2014-06-19 15:17 | rhansen | New Issue | |
| 2014-06-19 15:17 | rhansen | Name | => Richard Hansen |
| 2014-06-19 15:17 | rhansen | Organization | => BBN |
| 2014-06-19 15:17 | rhansen | Section | => c99, make |
| 2014-06-19 15:17 | rhansen | Page Number | => 2512, 2945 |
| 2014-06-19 15:17 | rhansen | Line Number | => 80970-80975, 97465 |
| 2014-06-19 16:51 | rhansen | Note Added: 0002269 | |
| 2014-06-19 18:10 | mdempsky | Note Added: 0002270 | |
| 2014-06-20 09:19 | joerg | Note Added: 0002271 | |
| 2014-06-20 09:20 | joerg | Note Edited: 0002271 | |
| 2014-06-20 09:26 | geoffclare | Note Added: 0002272 | |
| 2014-06-20 09:34 | joerg | Note Added: 0002273 | |
| 2014-06-26 15:48 | rhansen | Note Edited: 0002269 | |
| 2014-06-26 15:50 | rhansen | Note Edited: 0002269 | |
| 2014-06-26 15:51 | geoffclare | Interp Status | => Pending |
| 2014-06-26 15:51 | geoffclare | Final Accepted Text | => 0000848:0002269 |
| 2014-06-26 15:51 | geoffclare | Status | New => Interpretation Required |
| 2014-06-26 15:51 | geoffclare | Resolution | Open => Accepted As Marked |
| 2014-06-26 15:52 | geoffclare | Tag Attached: tc2-2008 | |
| 2014-06-27 11:09 | ajosey | Interp Status | Pending => Proposed |
| 2014-06-27 11:09 | ajosey | Note Added: 0002288 | |
| 2014-08-05 13:18 | ajosey | Interp Status | Proposed => Approved |
| 2014-08-05 13:18 | ajosey | Note Added: 0002332 | |
| 2019-06-10 08:54 | agadmin | Status | Interpretation Required => Closed |
