| View Issue Details [ Jump to Notes ] | [ Issue History ] [ Print ] | ||||||||
| ID | Project | Category | View Status | Date Submitted | Last Update | ||||
|---|---|---|---|---|---|---|---|---|---|
| 0000556 | mercury | Bug | public | 2022-04-06 12:10 | 2022-04-07 12:56 | ||||
| Reporter | juliensf | ||||||||
| Assigned To | zs | ||||||||
| Priority | normal | Severity | minor | Reproducibility | have not tried | ||||
| Status | resolved | Resolution | fixed | ||||||
| Product Version | |||||||||
| Target Version | Fixed in Version | ||||||||
| Summary | 0000556: Stage 2 compiler does not build with --intermod-opt -O5 and rotd-2022-04-04 | ||||||||
| Description | From testing.mercurylang.org fact_table.m:2543: error: the user defined inst fact_table.m:2553: error: the user defined inst fact_table.m:3400: error: the user defined inst building of stage 2 compiler not successful In non-truncated form it is: fact_table.m:2543: In the second argument of predicate `get_hash_table_type'/2: fact_table.m:2543: error: the user defined inst fact_table.m:2543: `ll_backend.fact_table.key_char'/0 fact_table.m:2543: is declared to be applicable only to values of type fact_table.m:2543: `char.char'/0, fact_table.m:2543: but here it is applied to values of type fact_table.m:2543: `character'/0. fact_table.m:2553: In the third argument of predicate fact_table.m:2553: `get_hash_table_type_loop'/3: fact_table.m:2553: error: the user defined inst fact_table.m:2553: `ll_backend.fact_table.key_char'/0 fact_table.m:2553: is declared to be applicable only to values of type fact_table.m:2553: `char.char'/0, fact_table.m:2553: but here it is applied to values of type fact_table.m:2553: `character'/0. fact_table.m:3400: In the fifth argument of predicate fact_table.m:3400: `generate_hash_lookup_code'/7: fact_table.m:3400: error: the user defined inst fact_table.m:3400: `ll_backend.fact_table.key_char'/0 fact_table.m:3400: is declared to be applicable only to values of type fact_table.m:3400: `char.char'/0, fact_table.m:3400: but here it is applied to values of type fact_table.m:3400: `character'/0. It looks like equivalence types are not being expanded in insts correctly. | ||||||||
| Tags | No tags attached. | ||||||||
| Attached Files |
| ||||||||
 Relationships |
|
 Relationships |
| Notes |
|
|
zs (developer) 2022-04-06 15:18 |
I started fixing this by expanding type_ctors in inst definitions, but realised that this is wrong approach. The reason is that the inst will (in virtually all cases) contain function symbols of the named type, and thus makes sense only if the user who wrote the inst knows the ultimate expansion of the "for typector". If they know it, why not write it? Therefore I propose to fix this bug in two ways. - Replace char with character in the inst definition in fact_table.m. - Add code that detects when the "for type_ctor" part of an inst defn is an equivalence type, and generates a version of the error messag above that mentions this fact. Any objections? |
|
juliensf (administrator) 2022-04-06 15:37 |
I don't have any particularly strong objections. Your proposed fix does introduce a certain lack of uniformity with respect to how equivalence types are treated by different parts of the language. (It's probably not too bad in presence of an error message describing what is wrong.) As to "if they know, why not write it?", I think this bug answers that question. It wasn't written because the equivalence type is the preferred name for the type in question. When deprecating a type name we quite often start by introducing an equivalence that defines the new name and then swapping them around later, before finally (after a respectable interval) deleting the original name. (That isn't quite the story with character / char, but it's close.) Another reason a user may have used char instead of character here, is that they may have read the library documentation. It is fairly emphatic on the point: Originally we used `character' rather than `char' for the type name because `char' was used by NU-Prolog to mean something different. But now we use `char' and the use of `character' is discouraged. |
|
zs (developer) 2022-04-07 12:56 |
Fix committed 2022 apr 7. |
| Notes |
| Issue History |
|||
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2022-04-06 12:10 | juliensf | New Issue | |
| 2022-04-06 12:22 | zs | Assigned To | => zs |
| 2022-04-06 12:22 | zs | Status | new => assigned |
| 2022-04-06 15:18 | zs | Note Added: 0001187 | |
| 2022-04-06 15:37 | juliensf | Note Added: 0001188 | |
| 2022-04-07 12:56 | zs | Status | assigned => resolved |
| 2022-04-07 12:56 | zs | Resolution | open => fixed |
| 2022-04-07 12:56 | zs | Note Added: 0001189 | |
| Issue History |