This topic is really poorly understood. Anyone who ever who tells you it's fine to just round up or round off when choosing spoke lengths hasn't been through the disaster scenarios that can be caused by it, or perhaps doesn't remember because the experience is too painful and embarrassing. Rounding up can cause you to run out of threaded length before the spokes are fully tensioned. Always round down unless you can test and measure the spokes in question before building to avoid running out of threaded length or poking the tube. If you have to order spokes, just round down. If your calculated optimal length is accurate (and there's no reason it shouldn't be) and you must round down to the next even number from there (as in the 292 in your example), nothing meaningful will ever be lost in terms of durability from that extra 1.99mm max.
What gets overlooked in this discussion is that there are whole brands/models of spoke, or spoke/nipple combinations, that offer zero ability for the spoke to protrude past the top of the nipple. They don't have the threaded length to do it. So rounding up walks you right into the failure situation of running out of thread and having to either start over or accept an untrueable wheel with bad tension. Various Wheelsmith bags I've used, for example, have been like this, with Wheelsmith nipples.
Contemporary DT Champions and Competitions, the most commonly encountered premium spokes, avoid this occurrence by having some superfluous/generous threaded length. Some other spoke brands and models do not, and give you no leeway for the spoke to be long. And single wall builds naturally have no such leeway.
Some point out that having aluminum nipples miss more than 1mm or so of engagement introduces some risk of the nipple breaking. That's probably true and I recommend shooting for the best possible thread engagement on aluminum nipples, which for practical purposes on conventional nipples is between the base of the screwdriver slot and the top. However, you'll never know if that's what caused it, because al nipples break anyway, especially for 2mm spoke ends. (Al nipples for 1.8s are far less prone to breaking because of the extra material.)
Your calculation must be accurate. You don't want to miss more thread engagement than around 2mm from perfect. Never use a calculator where you can't verify that the math it's doing is the pure classic spoke length formula. Some calculators try to invisibly factor in estimates of spoke stretch and rim compression, which you should do manually based on the parts you're using. (Spoke gauges differ in stretch, the amount of tension also causes it to differ, and the stoutness/cross-sectional area of rims make them differ in compression.) Because it's an Excel spreadsheet and all its formulas are right out in the open, Spocalc is my preference. And always measure the hub and rim yourself. It seems pedantic, but I promise that if you build enough wheels, or are ever working for a wheel producer and are tasked with doing spoke calculations and there's never time to do test runs of anything (this was my situation that taught me good spoke calculation habits), you'll find all this necessary to have consistently good results.
If you have to order the hub/rim along with the spokes (particularly the rim, because manufacturer list measurements for hubs are usually very accurate), just take vendor/manufacturer listed measurements on faith for calculation purposes, but measure yourself once you do have everything to make sure what you're about to do is going to work. Second-guess published ERDs early and often.
Wheelbuilding texts almost never remember to tell you that longer length nipples usually have more threads, and you need to compensate for this to avoid bottoming out the threads on the spoke. Most handbuilt wheels are done with 12mm nipples, but if you ever use 14s or especially 16s, you'll find the rim's ERD is effectively a few millimeters smaller. Or, in other words, ERD is usually calculated as the diameter of the circle formed by the tip of the spokes in a state of optimal thread engagement. Usually that's synonymous with the tip of the spoke being just about flush with the top of the nipple, somewhere between the bottom of the screwdriver slot and the top. 12mm nipples can all handle that, but most of the time if you thread an aftermarket/handbuilding-worthy/premium (however you want to say it) spoke into a 16, you'll find it runs out of thread and bottoms out before it reaches the top.
If you do need to round up, or insist on getting full thread engagement, always test it out first by taking a nipple and spoke that you're about to use and see how far the nipple screws on before it bottoms out, and then measure the amount of protruding spoke you have to work with to make sure it will be acceptable with whatever your calculated length was. Also make sure the rim's depth is such that it won't cause spoke to protrude up and poke the rim strip/tube. If the spoke meets these criteria it's fine, but it can't just be assumed that it will until you test. Do that test on enough spokes and you'll see.