IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
When there is an error introduced by #include-ing a file we should be able to
hit the disk (in cling's context) the next time. The user could fix it in the
meanwhile.
Newer clang supports a flag which can be used to mark the user files as volatile.
This means that the file stat will be invalidated, causing clang to hit the
disk and fetch the new content of the FileEntry.
The complication comes from the fact that when the file size of the file entry
is invalidated the cache in the SourceManager is not syncronized, thus clang
issued an error. The patch in clang checks if the file size == 0 and the
modification time == 0 this means that we are in cling's context and have to sync
the cache and continue with no errors.
This fixes an issue exposed by Jerome's implementation. Before we removed the
entire FileEntry to achieve the same behavior, however the SourceManager kept
the reference to it. This leads to seg faults when iterating over the included
files (eg. .files or .storeState)
71910784b3