Hi David, On 9/3/2025 3:04 PM, David Marchand wrote: > Hello, > > On Wed, 3 Sept 2025 at 04:05, Chengwen Feng <fengcheng...@huawei.com> wrote: >> >> Currently, the RTE_EXPORT_INTERNAL_SYMBOL, RTE_EXPORT_SYMBOL and >> RTE_EXPORT_EXPERIMENTAL_SYMBOL are placed at the beginning of APIs, >> but don't end with a semicolon. As a result, some IDEs cannot identify >> the APIs and cannot quickly jump to the definition. >> >> A semicolon is added to the end of above RTE_EXPORT_XXX_SYMBOL in this >> commit. >> >> And also redefine RTE_EXPORT_XXX_SYMBOL: >> #define RTE_EXPORT_XXX_SYMBOL(x, x) extern int dummy_rte_export_symbol >> >> Chengwen Feng (5): >> lib: add semicolon when export symbol >> lib: add semicolon when export experimental symbol >> lib: add semicolon when export internal symbol >> drivers: add semicolon when export any symbol >> doc: update ABI versioning guide > > I am skeptical about this series. > > The current positionning should be seen as an additional info on the > return type, in the definition of the symbol. > Does it mean that this IDE would fail if we add any kind of > macros/attribute involving the symbol name?
I tried vscode and SI (source insight), and found user could use "token macro" in SI to make the IDE skip such symbol (so use ctrl+ will quick jump to definition), but I can't find such setting for vscode. > > Afaics, ctags can be taught to skip those macros and just behaves > correctly by adding in its config file: > -DRTE_EXPORT_EXPERIMENTAL_SYMBOL(a)= > -DRTE_EXPORT_INTERNAL_SYMBOL(a)= > -DRTE_EXPORT_SYMBOL(a)= How about add note in DPDK document if don't apply this commit? > > I think another option would be to move the call to export macros > after the whole definition of the symbol, though I prefer the current > position for readability. If not add a semi, it will affect next API: int A() RTE_EXPORT_SYMBOL(A) int B() > >
