I'm uncertain about the length of the u in the perfect and perfect passive participle stems of the verb uro /uːroː/.
My research
Lewis (1890) gives "ūrō ūssī, ūstus" but doesn't explain why.
On the other hand, Lewis and Short leaves the vowel unmarked "ūro , ussi, ustum", which doesn't indicate any specific length.
The Wiktionary entry for ūrō gives "perfect active ussī, supine ūstum", also with no explanation. (Unmarked vowels in Wiktionary are supposed to be interpreted as short, because unlike L&S, Wiktionary does not use breves.)
De Vaan gives "ūrō" "ussī" and "ustum". I feel like I've seen De Vaan leave off macrons in closed syllables in some places, but in this case I am fairly sure that De Vaan really thought of the vowel as short, because a separate PIE form "*usto- 'burnt'" is given for the past participle (vs. the etymology "ouse/o- 'to burn'" given for the stem of the forms with a long vowel).
I don't know enough about PIE morphology to understand whether an alternation between *ou and *u, which looks like ablaut, would be plausible in this context.
Based on IE comparison, a short vowel seems to be expected in the p.p. participle, but not in the perfect
Wiktionary gives the PIE root as *h₁ews- (a form that agrees better with what I recall of a theorized phonotactic constraint against word-initial vowels in PIE). The Wiktionary entry for *h₁ews- does agree with De Vaan in showing a short u in the participle: the PIE form is given by Wiktionary as *h₁us-tós, yielding Sanskrit "उष्ट (uṣṭá)" as well as Italic *ustos.
However, it lists Latin "ussī" as a descendant of a supposed "s-aorist" form "h₁ḗwst ~ *h₁éws-n̥t", which is also supposed to be the source of Ancient Greek εὗσα, the aorist form of εὕω. As far as I can tell, neither the Ancient Greek form, nor the PIE aorist form that Wiktionary gives is a zero-grade form, so the alleged short u in Latin ussi does not seem to be explained by the etymology given in the Wiktionary article for *h₁ews-. I would expect PIE *ew to become Latin ū (I also can't think of a Latin sound change that would shorten ū to u in this context).
Leveling of vowel length between the perfect and perfect participle doesn't seem implausible (I forget the details of how these forms developed in Latin, but I remember learning that they influenced each other by analogy in at least some cases) but I don't know how we could know that leveling occurred for this verb, or that the short rather than the long vowel was leveled. To me, it also seems plausible to suppose that the past participle could have gained a long vowel by analogy with the present stem.
I was looking at the length of the vowel in some other verbs, and I saw that in iubeo, De Vaan says that leveling is supposed to have resulting in a short vowel being generalized from the perfect passive participle to both the perfect and the present stems.
Romance evidence doesn't seem clear
I haven't seen any evidence from the Romance languages that clearly points towards a short vowel in either form. Italian comburere has re-formed both the perfect (to comburei) and the past participle (to comburuto). There seem to be Romance reflexes of a derived verb ustulo, and the vowels I see in Italian "ustolare, brustolare" and French "brûler" look more consistent with a long than a short quantity in the Latin etymon.
Questions
What was the length of the u in ussi and ustus, and how do we know? More specifically, what reasons, if any, are there for thinking that ussi had a short vowel?