Add other Julia compilation flags

[lucatrv]

No description provided.

[lucatrv]

@SimonDanisch @NHDaly
this is a quite extensive PR, please review it, if you agree I will finalize and merge it.

[NHDaly]

I'm interested in the use case for disabling/enabling --precompiled and --compilecache. Is there a reason that a user of this package would want that? If so, could we maybe also explain that a bit in the docstring?

[NHDaly]

From the arg name, "--startup-file", I would assume that this flag allows me to set a custom startup file. Is that desirable? I would either consider changing the behavior to what I just described, or changing the name of the variable to something like --use-startupfile?

[lucatrv]

As explained below, juliac provides same compilation flags as julia (in future julia-compile). See julia -h.

[NHDaly]

okay i'm convinced. :) thanks!

[NHDaly]

(sorry i misunderstood that what you were doing in this PR is just copying forward all of the julia flags! Makes sense.)

[NHDaly]

"these" is unprecise. How about:

#  TODO: precompiled and compilecache may be removed, pending: https://github.com/JuliaLang/PackageCompiler.jl/issues/47

[lucatrv]

Ok I'll change that.

[NHDaly]

hmm, i could be misunderstanding, but the first one doesn't feel correct to me.
If the user passes precompiled = "yes", won't this set precompiled to "no"?

I think maybe you want parentheses around the second two phrases?

 precompiled == nothing && (cpu_target == nothing || (precompiled = "no"))

But honestly that's still hard to parse. perhaps this can simply be an if-statement? I think this is what you meant?:

if (precompiled == nothing && cpu_target != nothing) precompiled = "no" end
# or this:
(precompiled == nothing && cpu_target != nothing) && (precompiled = "no")

[lucatrv]

In Julia (like in Bash) && is passed through only if the previous expression returns true, while || is passed through only if the previous expression returns false. So if precompiled = "yes", then precompiled == nothing returns false, and therefore && is not passed through.

[NHDaly]

:) So I think everything you just said is correct, and yet, behold:

julia> false && true || println("HI")
HI

I think the nuance is here:

while || is passed through only if the previous expression returns false.

In this case, the previous expression is actually the truth-value on the left-hand side of the ||. I would imagine it like this:

`precompiled == nothing` --> <evaluate expression ... okay, it's false!> --> `false` --> `&& (cpu_target == nothing)` --> <okay, skip the right-hand side, cause this already evaluates to false --> `false` --> `|| (precompiled = "no")` --> <now we need to evaluate the right-hand side again because the left-hand side was false, so the truthiness of the whole expression depends on the right-hand side> --> (precompiled = "no") --> <returns `nothing`>.

That's why I suggested, if you want to keep it exactly as you currently have it, you need to put parens around the entire right-hand side of &&, so that it can skip the whole thing:

julia> false && (true || println("hello?"))
>> false

But i still think an if-statement might be better, since, clearly, multiple short-circuiting statements can be confusing and hard to read. :)

[NHDaly]

(In bash as well):

$ false && true || echo "HI"
HI
$ false && (true || echo "helloooooooooo")
$

[lucatrv]

You're perfectly right of course :) it's due to the precedence of && over ||, I'll fix that, thanks!

[lucatrv]

This is pretty straightforward:

precompiled == nothing && cpu_target != nothing && (precompiled = "no")

[NHDaly]

👍 Otherwise, this looks good to me!

[lucatrv]

I'm interested in the use case for disabling/enabling --precompiled and --compilecache. Is there a reason that a user of this package would want that?

We are still in an early stage of development of Static Julia, there are still doubts for instance whether disabling compilecache is required just because of a julia bug or other. Moreover things may change once julia-compile is introduced, which may provide behaviors specifically addressed at static compilation.
So my intent is for now to provide full testing capabilities, and to keep juliac aligned as much as possible with julia (now) and with julia-compile (in future), so that it is possible to verify all possible flags combinations, while keeping safe defaults.

If so, could we maybe also explain that a bit in the docstring?

I agree that in future, once everything is clarified, we could add explanations in the docstring, but for now I thought to leave juliac docstring as aligned as possible with julia, so that we can easily follow its development changes, until everything is clarified.

[codecov-io]

Codecov Report

Merging #62 into master will decrease coverage by 0.29%.
The diff coverage is 78.57%.

@@            Coverage Diff            @@
##           master      #62     +/-   ##
=========================================
- Coverage   69.83%   69.53%   -0.3%     
=========================================
  Files           5        5             
  Lines         295      302      +7     
=========================================
+ Hits          206      210      +4     
- Misses         89       92      +3

Impacted Files	Coverage Δ
src/snooping.jl	78.94% <ø> (ø)	⬆️
src/api.jl	28.57% <ø> (ø)	⬆️
src/static_julia.jl	75.36% <78.57%> (-0.98%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 60a4c6a...8c72883. Read the comment docs.

[NHDaly]

👍👍👍

…

On Sat, May 12, 2018, 9:30 AM Luca Trevisani ***@***.***> wrote: Merged #62 <#62>. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#62 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABgkEawhY0ZHizacMDyELmmgKMf270LKks5txvH6gaJpZM4T6gpy> .