On Sat, Mar 06, 2021 at 06:07:43PM +0000, Thorsten Glaser wrote: > Sven Joachim dixit: > > >I see that this might be a problem (albeit unlikely to happen in > >practice), however I have trouble understanding exactly where a > >use-after-realloc bug comes into play. Maybe Thorsten can help me fix > >my blindness? > > The next time something is selected, the code a little further > up will check if the allocated size is sufficient, and, if so, > use screen->selection_data which was the pre-realloc address of > line. > > >> I am glad and surprised that sid is okay and there doesn't seem to be > > The code in sid completely differs (structures, variable names, etc). The renaming (selection_size) comes from patch #338, which looks like this item: Patch #338 - 2018/12/09 * amend solution for Debian #758633 to ensure that replies for bracketed paste are not sent while processing a selection for exec-formatted (Debian #913237). > >suggestion you could also apply the patches to the SaltTextAway() > >function from xterm 365e. > > If 365e is like 366 (currently in sid), you’ll have lots of fun due > to the renamed everything. 366 is current. I have some changes for 367 which I'll put out after seeing what I can do to improve performance with fwvm active-icon. > I’d rather Tom changed xterm upstream to address the realloc-failure > difference. I know he reads Debian bugreports ;-) and he’s really > busy so probably takes longer to respond. it used to be the case that downstream would ask my opinion on patches like this -- it's been a while since anyone did -- Thomas E. Dickey <dickey@invisible-island.net> https://invisible-island.net ftp://ftp.invisible-island.net
Attachment:
signature.asc
Description: PGP signature