Bug 8765 – assert should print the source code for the condition when no message argument present

I'd love to see the following: assert(5 == 4); print out the actual source code of the condition on failure, that is "[email protected](123): 5 == 4". This should happen when there is no user-defined message (and maybe _in addition to_ the provided message). Currently I just get "unittest failure" instead of the condition source, which is useless. This would make it far, far easier to track down which assert failed without having to actually go look at the line number in the file. Also, this is what most unit-testing libraries for other languages do already, like for example GoogleTest for C++ etc. Since assert() is not a function but an expression in the language, this should not be impossible to implement, should it? GoogleTest does it with macros, but (thank God) we don't have those in D.

(In reply to comment #0) > assert(5 == 4); When there is no message I can change this to: [email protected](5): assert(5 == 4) Would this be ok with Walter? It's a 2 line change in the front-end.

(In reply to comment #1) > (In reply to comment #0) > > assert(5 == 4); Actually wait a minute, it already does this in 2.060. Is there some other example where this doesn't work?

(In reply to comment #2) > Actually wait a minute, it already does this in 2.060. Is there some other > example where this doesn't work? If I compile and run this program (Windows): void main() { assert(5 == 4); } It gives me: core.exception.AssertError@test(2): Assertion failure Followed by a stack trace.

(In reply to comment #3) > (In reply to comment #2) > > > Actually wait a minute, it already does this in 2.060. Is there some other > > example where this doesn't work? > > If I compile and run this program (Windows): > > > void main() { > assert(5 == 4); > } > > > It gives me: > > core.exception.AssertError@test(2): Assertion failure > > Followed by a stack trace. OK will fix then.

Same as bearophile; the following program void main() { assert(5 == 4); } gives me the following output: core.exception.AssertError@test(2): Assertion failure ---------------- 5 test 0x000000010d95b05a _d_assertm + 38 6 test 0x000000010d948df7 void test.__assert(int) + 23 7 test 0x000000010d948dda _Dmain + 14 8 test 0x000000010d95b9ae extern (C) int rt.dmain2.main(int, char**).void runMain() + 34 9 test 0x000000010d95b365 extern (C) int rt.dmain2.main(int, char**).void tryExec(scope void delegate()) + 45 10 test 0x000000010d95b9f8 extern (C) int rt.dmain2.main(int, char**).void runAll() + 56 11 test 0x000000010d95b365 extern (C) int rt.dmain2.main(int, char**).void tryExec(scope void delegate()) + 45 12 test 0x000000010d95b2ef main + 235 13 libdyld.dylib 0x00007fff89dae7e1 start + 0 14 ??? 0x0000000000000001 0x0 + 1 ---------------- My system/configuration is as follows: OS: Mac OS X Mountain Lion 10.8 DMD64 v2.060 rdmd build 20120724 command: rdmd -unittest test.d

*** Issue 8058 has been marked as a duplicate of this issue. ***

Time for a resurrection?

(In reply to comment #8) > Time for a resurrection? Yeah, I'll give it another go soon.

(In reply to bearophile_hugs from comment #3) > It gives me: > > core.exception.AssertError@test(2): Assertion failure > > Followed by a stack trace. I don't get it - what's the problem?

(In reply to Val Markovic from comment #0) > Currently I just get "unittest failure" instead of the condition > source, which is useless. It gives the file/line, which is not useless: 1. editors, IDEs, etc., can take you right to the source code of the error 2. assert checks can add a great deal of bloat to the executable - the current implementation tries to minimize that. Adding expression strings to all of them will dramatically increase the size of an executable I just do not understand why file/line is not sufficient.

(In reply to Walter Bright from comment #11) > (In reply to Val Markovic from comment #0) > > Currently I just get "unittest failure" instead of the condition > > source, which is useless. > > It gives the file/line, which is not useless: > > 1. editors, IDEs, etc., can take you right to the source code of the error > > 2. assert checks can add a great deal of bloat to the executable - the > current implementation tries to minimize that. Adding expression strings to > all of them will dramatically increase the size of an executable > > I just do not understand why file/line is not sufficient. If the assert is triggered at runtime in a realtime application you want to make sure you get the relevant information back to you as soon as possible. You may not even have the source at hand, it would really help knowing what condition failed. Code bloat is an issue, but if the requirement is a minimal binary then it would be safe to assume -release is implied, no? So the asserts wouldn't be compiled in anywho. In any case I'll work on this PR again.

my PR solves exactly this problem but in a more general way: https://github.com/dlang/dmd/pull/7821

Meanwhile `-checkaction=context` exists which will print the failing expression using the actual values instead of the plain source code. Can this be closed?

Yes, this has been fixed by: https://github.com/dlang/dmd/pull/8517