1# bmcweb headers 2 3**Why does bmcweb use so many headers? My build times are slow!** 4 5TL;DR, History 6 7bmcweb at one point was a crow-based project. Evidence of this can still be seen 8in the http/.hpp files that still contain references to the crow namespaces. 9Crow makes heavy use of headers and template meta programming, and doesn't ship 10any cpp or implementation files, choosing to put everything in include once 11headers. As bmcweb evolved, it needed more capabilities, so the core was ported 12to Boost Beast, and what remains has very little similarity to crow anymore. 13Boost::beast at the time we ported took the same opinion, relying on header 14files and almost no implementation compile units. A large amount of the compile 15time is taken up in boost::beast template instantiations, specifically for 16boost::beast::http::message (ie Request and Response). 17 18The initial solution that gets proposed is to just move everything as it exists 19to separate compile units, making no other changes. This has been proposed and 20implemented 3-4 times in the project, the latest of which is below. The intent 21of this document is largely to save effort for the next person, so they can at 22least start from the existing prior attempts. 23 24<https://gerrit.openbmc.org/c/openbmc/bmcweb/+/49039> 25 26Moving to cpp files without handling any architecture has the net result of 27making total compilation slower, not faster, as the slowest-to-compile parts end 28up getting compiled multiple times, then the duplicates deleted at link time. 29This isn't great for the end result. 30 31To actually effect the result that we'd like to see from multiple compile units, 32there have been proposed a few ideas might provide some relief; 33 34- Moving the Request and Response containers to opaque structures, so a majority 35 of code only needs to #include the interface, not any of the template code. 36 <https://gerrit.openbmc.org/c/openbmc/bmcweb/+/37445> Doing this exposed a 37 number of mediocre practices in the route handlers, where routes made copies 38 of requests/responses, relied on APIs that should've been internal, and other 39 practices that make this migration less straightforward, but is still being 40 pursued by maintainers over time. 41- Moving the internals of Request/Response/Connection to rely on something like 42 [http::proto](https://github.com/CPPAlliance/http_proto) which, written by the 43 same author as boost::beast, claims to have significant reduction in compile 44 time templates, and might not require abstracting the Request/Response 45 objects. 46- Reduce the bmcweb binary size to the point where link time optimization is not 47 required for most usages. About half of the bmcweb build time is spent doing 48 link time optimization, which, as of this time is required to keep bmcweb code 49 small enough to deploy on an actual BMCs (See DEVELOPING.md for details). One 50 could theoretically determine the source of where LTO decreases the binary 51 size the most, and ensure that those were all in the same compile unit, such 52 that they got optimized without requiring LTO. 53