Fix infinite loop from #21161 - repr of NumberFields (the parents) should indicate its embedding if there is one

[chapoton]

Because this is causing some random infinite loops when running the testsuite. To reproduce, as reported in ​https://github.com/cschwan/sage-on-gentoo/issues/541, use:

sage -t --long src/sage/structure/ src/sage/interfaces/

Another symptom:

The doctest

File "src/sage/structure/parent.pyx", line 1734, in sage.structure.parent.Parent.hom.register_embedding
Failed example:
    K.coerce_embedding()(a)

when it does not fail, and then one calls

K.coerce_embedding()

again, makes sage crash.

Removing the change of repr made in #21161 fixes that.

For the complete log when the doctest fails, see for example

​https://patchbot.sagemath.org/log/27408/Ubuntu/18.04/x86_64/4.15.0-52-generic/petitbonum/2019-06-20%2014:50:38?short

[chapoton]

Adding the single line

gen = self.gen_embedding()

is enough to trigger the crash.

[chapoton]

This may be considered as a blocker, no ?

[gh-timokau]

Thank you for analyzing this issue! Previous discussion in ​https://github.com/cschwan/sage-on-gentoo/issues/541, were we assumed it was distro related.

I agree that it should be included in 8.8.

[jipilab]

Replying to chapoton:

Adding the single line

gen = self.gen_embedding()

is enough to trigger the crash.

I do not get a crash with 8.8.rc2... :-S Hmm.

So there is no way to bypass the current issue and still be able to know via the repr that the NumberField? is embedded? This is in itself problematic, right?

Reverting completely #21161 would be too bad... Isn't there any way to remove the crash caused by calling

gen = self.gen_embedding()

?

[chapoton]

Nota Bene: I got the crash/not crash when using py3-sage. Not tried with py2-sage.

[jipilab]

Replying to chapoton:

Nota Bene: I got the crash/not crash when using py3-sage. Not tried with py2-sage.

Hmm. I also tested with py3-sage. But I might have some obscure side effects...

[chapoton]

Here is a branch that fixes the crash.. Not sure if this is the right thing to do. And no idea if this prevents the random infinite loop to re-appear.

---

New commits:

​c6c70e7

trac 28036 fix proposal

[chapoton]

the fix proposal provokes a few failing doctests, that may not be so simple..

[chapoton]

Here is another proposal.. (not really convincing either)

---

New commits:

​fe00485

another tentative fix

[mmezzarobba]

Replying to chapoton:

Here is another proposal.. (not really convincing either)

Did you test it? I don't understand how it can work, since _embedding is a cdef attribute. Apart from that, the idea looks sensible enough to me.

Another option may be to use a lazy string for the error message that causes the loop.

[mkoeppe]

Replying to mmezzarobba:

Another option may be to use a lazy string for the error message that causes the loop.

+1

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​c44fd16

Use LazyFormat to fix infinite loop

[mkoeppe]

I was able to reproduce this error. I agree it's a blocker. Using LazyFormat seems to fix it for me. Please test.

[mkoeppe]

(Sorry for overwriting your branch.)

[chapoton]

This is a better solution indeed. Works fine for me (on py3-sage).

[dimpase]

can that 1-line way to crash above be put into a doctest?

[mkoeppe]

Replying to dimpase:

can that 1-line way to crash above be put into a doctest?

I don't know how to reproduce the crash with a single doctest.

[chapoton]

add one line

K.coerce_embedding()

after the line

File "src/sage/structure/parent.pyx", line 1734, in sage.structure.parent.Parent.hom.register_embedding
Failed example:
    K.coerce_embedding()(a)

[mkoeppe]

This triggers the error for me in an interactive session; but not if I add it to the doctest...