Changes between Initial Version and Version 1 of Ticket #16995, comment 13
- Timestamp:
- 2018-11-19T07:33:03+01:00 (7 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #16995, comment 13
initial v1 10 10 ----- 11 11 12 I do support using the unsigned int API for that other pro plem as an immediate measure, but the long term fix is using a long data type. I've checked {{{time_t}}} on x86_64 Linux to be a datatype of 8 bytes length ({{{long int}}}) already, dictated by Linux headers.12 I do support using the unsigned int API for that other problem as an interim, but the long term fix is using a long data type. 13 13 14 The notable exception is that if you compile 32 bit code on x86_64 it pulls declaration as to be 4 bytes wide ({{{int}}}) from the headers. You can check this yourself - there is a tiny prog and details on how to compile it on the [https://de.wikipedia.org/wiki/Diskussion:Jahr-2038-Problem#Code,_um_Aussagen_des_letzten_Abschnitts_zu_pr%C3%BCfen german wikipedia talk page on the problem] that exploits the details. 14 {{{time_t}}} on x86_64 Linux is, by default, a datatype of 8 bytes length ({{{long int}}}) already, dictated by Linux headers. 15 The notable exception occurs when code is compiled as 32 bit binary on x86_64 using e.g. {{{gcc -m32}}}. 15 16 16 But again, we should move that to another bug for this (or rename this one) as it has another scope and broader implications than the original issue here (fixed by the smaller of attached patches).17 In the latter case headers will declare {{{time_t}}} to be 4 bytes wide. You can check this yourself, see e.g. the [https://de.wikipedia.org/wiki/Diskussion:Jahr-2038-Problem#Code,_um_Aussagen_des_letzten_Abschnitts_zu_pr%C3%BCfen german wikipedia talk page on the problem] that has c code to exploit the details. 17 18 19 But again, we should move this issue to another ticket. It has a much broader scope than the OP's issue and can be dealt with separately - and it's less urgent than the issue here.


