Bug 217149 - seq(1) inconsistently omits 'last' when using float increment

[yuripv_icloud.com]

Summary

Don't accumulate rounding errors:

loki:yuri:~/ws/pr217149/usr.bin/seq$ seq 1 .1 2
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
loki:yuri:~/ws/pr217149/usr.bin/seq$ /usr/obj/home/yuri/ws/pr217149/amd64.amd64/usr.bin/seq/seq 1 .1 2
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
loki:yuri:~/ws/pr217149/usr.bin/seq$ seq 1.6 .05 2
1.6
1.65
1.7
1.75
1.8
1.85
1.9
1.95
loki:yuri:~/ws/pr217149/usr.bin/seq$ /usr/obj/home/yuri/ws/pr217149/amd64.amd64/usr.bin/seq/seq 1.6 .05 2
1.6
1.65
1.7
1.75
1.8
1.85
1.9
1.95
2

[imp]

Someone had a problem and thought solving it with float was a good idea. Then we had 1.999999 problems. :)
This fixes it to use the right calculation that doesn't accumulate error.

[imp]

usr.bin/seq/seq.c

189–191

Now that I look at this, I'm thinking it might be better to compute an exact number of steps and just do a simple for loop with step...

for (step = 0; step < max_step; step++)

190

since step starts at one, doesn't this skip the first step?

[yuripv_icloud.com]

usr.bin/seq/seq.c

189–191

I thought about it, but it doesn't make this any more readable, at least in the way I was able to implement it.

190

No, on first iteration we print cur set to first, and only then compute cur using step being 1.

[yuripv_icloud.com]

fix the loop

[yuripv_icloud.com]

usr.bin/seq/seq.c

190

Actually, that was good question, initial revision would break seq 1 2 2 which should print only 1, but did print 1 3, updated.

[yuripv_icloud.com]

Add basic tests for fp increment.

I have added Kyle to reviewers in case he could look if tests are correct..

[kevans]

usr.bin/seq/tests/seq_test.sh

51

Can you split these two sanity checks out into another test, so we don't conflate what's actually a floating point test with what's a test of standard operation? The supported workflow for running tests is to make check or run with kyua, which will run all of the tests, so it's better to structure our tests accordingly.

[yuripv_icloud.com]

usr.bin/seq/tests/seq_test.sh

51

Done, and thanks for the handholding, it's all a bit new for me.

[yuripv_icloud.com]

updated copyright in test script as noted by Kyle.

[yuripv_icloud.com]

update to fix the test case Jilles provided -- compare the string representation as an additional check we didn't just hit rounding error; add relevant test case.

[pstef]

This was fixed by 3049d4cc (r330086).